Unlock a Bot. Come back to finished work.

Turns fuzzy ideas into shippable PRDs, user stories, and acceptance criteria a team can build against without a follow-up meeting.

Plans customer interviews, writes non-leading scripts, and synthesizes notes into opportunities, jobs-to-be-done, and kill criteria.

Prioritizes bets against capacity, sequences releases, and keeps the roadmap honest when everything is suddenly P0.

Turns engineering updates into crisp sprint reviews, stakeholder notes, and a single paragraph a CEO will actually read.

Defines north-star and input metrics, writes experiment briefs, and reads results without celebrating noise.

Runs go-to-market checklists, release comms, support enablement, and launch-day war rooms so shipping is not a surprise.

Helps run 1:1s, career conversations, and tough feedback with a clear bar and no theater.

Challenges design docs, tradeoffs, and org-level technical bets the way a sharp staff-plus engineer would in review.

Finds what is actually blocking the train and rebuilds the operating cadence: rituals, WIP limits, and ownership.

Runs severity calls, stakeholder comms, and postmortems that change the system instead of assigning blame.

Writes interview rubrics, structures loops, and calibrates engineering offers so the bar does not drift.

Walks a feature through STRIDE and data-flow review and returns a concrete threat list with mitigations, not a scare deck.

Reviews designs and diffs for authz gaps, injection, secrets handling, and supply-chain risk before they ship.

Triages CVEs, SBOMs, and upgrade risk so you patch what is reachable and stop treating every advisory as a fire drill.

Finds leaked credentials, unsafe logging, and sloppy key handling in code, configs, and CI, then tells you how to rotate and lock it down.

Installs lightweight security gates into the workflow you already have: review checks, threat-model triggers, and release gates.

Reviews IAM, network, and storage posture for common cloud footguns and returns a prioritized hardening list.

Runs lawful due diligence from public records, filings, and official registries. No dark-web dumps, no access tricks.

Tracks public mentions, impersonation, and leaked brand assets across the open web and social surfaces.

Builds competitor snapshots from public sites, filings, job posts, and social signals, cited to primary sources.

Maps an organization's public digital footprint: domains, certificates, and exposed services that are already on the open internet.

Verifies claims against primary public sources, flags weak citations, and rewrites briefs so every assertion has a URL.

Writes and reviews ESP-IDF and Arduino firmware for sensors, power modes, and OTA without cargo-cult samples.

Designs Meshtastic and LoRa topologies, channel plans, and node roles for coverage, hops, and battery life.

Translates schematics and pinouts into firmware tasks, bring-up checklists, and hardware-in-the-loop test plans.

Plans provisioning, OTA rollout, and remote diagnostics for small ESP32 fleets that cannot afford a brick.

Squeezes sleep current, wake sources, and battery life on ESP32 designs and tells you which milliamp is a lie.

Turns noise into a one-page weekly brief: risks, bets, cash, and the two asks that actually need the CEO.

Drafts board updates, the narrative arc, and Q&A for the room you actually have, not a generic deck.

Shapes fundraising narrative, use of funds, and investor follow-ups so the story survives the first skeptical partner.

Forces decisions into write-ups: options, the bet, kill criteria, and who owns the next check-in.

Owns the technical north star, kills accidental complexity, and says no to the fifth way of doing the same thing.

Runs make, buy, or partner calls with cost, lock-in, talent, and time-to-value on the table.

Decides what becomes a platform versus a one-off, and when extracting a platform is still too early.

Designs the engineering brand, ladder, and hiring story a CTO can stand behind in a closing conversation.

Balances bets across squads, kills zombie products, and funds the work that compounds.

Stress-tests packaging, willingness to pay, and migration paths so pricing is a product decision, not a spreadsheet accident.

Shapes squads, rituals, and the PM quality bar as the org scales past heroics.

Runs CAB agendas and turns executive customer signal into portfolio moves instead of feature souvenirs.

Reviews API shapes, error contracts, and versioning so clients do not inherit your internal mess.

Reads schemas and slow queries, then proposes indexes, access patterns, and the one rewrite that is actually worth it.

Audits UI for accessibility, keyboard paths, and semantic markup, then files fixes a designer and engineer can share.

Owns cut lists, changelog discipline, and rollback plans so a release is boring on purpose.

Designs growth experiments with a hypothesis, guardrails, and a readout that does not move the goalposts after the fact.

Writes search-ready briefs, titles, and internal-link plans grounded in intent, not keyword stuffing.

Turns a prospect's stack into a demo path, objection map, and a technical win plan the AE can run.

Triages hairy tickets, drafts customer-safe updates, and writes the internal RCA the product team can act on.

Builds search strings, outreach that does not sound like spam, and scorecards for roles you are actually trying to fill.

Turns warehouse questions into queries, charts, and a plain-language answer a PM or exec can use the same day.

Turns messy notes into owners, due dates, and a follow-up you can paste into Slack without rewriting it.

Runs the live deal: ZoomInfo for the buying committee and intent, Gong for the last call, Outreach or Amplemarket for the next touch, then writes the truth back to Salesforce and HubSpot.

Builds named lists from ZoomInfo, enriches through Clay, scores Hot/Warm/Cold, then loads Amplemarket and Outreach sequences. CRM write-back only after a real person is attached.

Produces a one-page account brief: ZoomInfo firmographics, intent, scoops and tech stack, Clay research, Gong last-call summary, plus the Salesforce and HubSpot record as it actually sits today.

Maps decision-makers with ZoomInfo, fills gaps with Clay and Amplemarket people search, flags missing titles in Salesforce and HubSpot, and recommends the next Outreach thread.

Pulls Gong account summaries, deal insights, and call briefs, then maps risk language to the Salesforce opportunity and HubSpot deal. Flags stalled next steps, not vibes.

Keeps contacts, companies, deals, and tickets in sync between Salesforce and HubSpot. Enriches empty fields from ZoomInfo and Clay. Never invents a close date.

Reads Salesforce opportunities and HubSpot deals, overlays Gong deal insights, and produces a forecast note: commit, upside, slip risk, and the one move that changes the week.

Builds meeting prep from ZoomInfo, the Gong call history, Outreach Kaia meetings, and the CRM record. Hands the AE a one-pager: who is in the room, last commitments, open tickets, suggested ask.

Uses ZoomInfo TAM sizing, similar-company search, and tech-stack snapshots, then Clay research, to turn an ICP into a named territory. Writes the resulting account universe into Salesforce and HubSpot.

Watches HubSpot tickets and Salesforce records for expansion and churn risk, pulls Gong account summaries, and recommends a multi-thread via ZoomInfo plus an Outreach follow-up.

Writes 1-3 ZoomInfo-personalized email variants, grounded in Clay research and the last Gong brief. Does not send. Hands the copy to Outreach or Amplemarket for the human to fire.

Audits Salesforce and HubSpot objects, sequence coverage in Outreach and Amplemarket, and Gong insight gaps. Returns a hygiene and coverage list a RevOps lead can act on the same day.

Finds B2B advertisers in Meta Ad Library and Google Ads Transparency, takes phones only from official sites and ad creatives, and stages a CRM list for the owner to fire. Does not send.

Finds editorial pages that AI answers already cite for buying prompts, keeps the repeats, and drops Reddit and other UGC. Does not email.

Finds editor or owner contacts on the official site, masthead, About, and public whois for a cited editorial URL. Caps at five contacts per domain. Does not email. Does not invent addresses.

Writes 80–120 word mention-pitch emails (mention swap, editorial backlink, or flat fee) for a cited page. Does not send. Hands the draft to the owner.

Triages inbound editor replies the owner pastes, drafts the mention copy and a fee card, and never pays. Owner sends and spends.

Rechecks agreed URLs for the live brand mention and whether category answers still cite the page. Does not email.

Reads the week's scouted pages, drafts, placements, and open asks, then recaps a Monday digest in chat. Never sends. Never spends.

Ships iOS and the shared mobile layer with screenshot proof, then watches the PR until checks are green. Does not merge.

Runs the desktop client change through a cloud agent, reads failing checks, and stages a proof-backed PR. Does not merge. Does not own the release train.

Takes infrastructure tickets with no owner, reproduces from logs, starts a cloud agent, and writes a cited handoff. Does not page people. Does not merge.

Ships Android client work with emulator or screenshot proof, then watches the PR until checks are green. Does not merge.

Tightens agent prompts, skills, and routines, and packages what a cloud agent learned. Does not merge product clients.

Keeps engineering playbooks current and runs a weekday alignment recap in chat. Does not write application code. Does not merge. Does not page.

Builds an hourly attention list from Slack the owner scopes, then filters noise against it. Mail or calendar only if the owner pastes them. Drafts only. Owner sends and archives.

PM-facing eng manager: decomposes goals, delegates to engineering or cloud agents, and checks output against the goal. Does not write the patch. Does not merge.

Turns Slack threads and pasted meeting notes into RFCs, research briefs, and product docs. Draft only. Owner edits and sends.

Morning or tactical product pulse from CSV, SQL results, or dashboard exports the owner pastes. Charts plus short synthesis. Does not connect to a warehouse. Does not invent metrics.

Turns product notes into a one-page GTM launch brief: audience, promise, channels, objections, and success metrics. Draft only. Does not post or buy ads.

Drafts a customer story or case study from call notes and metrics the owner pastes. Does not email the customer. Does not publish.

Builds and refreshes support macros from ticket themes the owner pastes. Draft only. Does not send replies to customers.

Triages a pasted ticket export into severity buckets and next owners. Pasted CSV only — not live Slack escalation. Does not reply to customers. Does not close tickets.

Scores inbound form or waitlist leads from a CSV the owner pastes against ICP rules. Does not email. Does not add to a sequence.

Drafts referral ask messages from closed-won notes the owner pastes. Does not send. Owner sends while present.

Builds a vendor renewal calendar and draft chase notes from contracts or invoices the owner pastes. Does not pay. Does not send.

Turns a messy how-to the owner pastes into a short ops runbook. Draft only. Does not change production systems. Slack is not packed — paste the thread.

Drafts interview scorecards and debrief memos from role requirements and interviewer notes the owner pastes. Paste-only debrief — not a hiring-bar clone. Does not email candidates.

Drafts offer letters and compensation breakdowns from salary bands and role details the owner pastes. Does not send the offer. Not legal advice.

Builds a cohort retention table and short readout from an events or user-activity CSV the owner pastes. Retention cohorts only — not a general pulse or warehouse Q&A. Does not invent retention rates.

Turns pasted A/B experiment stats into a one-page ship/hold/kill readout. Pasted-stats readout only — does not design the experiment. Does not invent p-values. Does not change feature flags.

Coordinates agents in separate worktrees, tracks file ownership, and proposes a safe merge order. Never merges or force-pushes. Humans keep the final merge.

Compares product manuals, usage patterns, and evidence to separate legitimate failure from misuse. Always recommends human review before any claim decision. Never files a claim alone.

Turns a recent git log or status the owner pastes into next-day priorities and tasks. Does not connect to GitHub. Never commits or pushes.

Highlights ambiguous contract clauses and dispute risk before signature. Always requires legal counsel review. Not a lawyer. Never signs or negotiates alone.

Sorts owner-pasted GitHub issues or bug lists by severity and type into a Fix/Test queue with draft labels and triage notes. Never merges, closes, or labels issues unattended.

Takes safe-to-patch items (flaky tests, lint) from pasted context and drafts patch notes plus PR description text for the human to open. Never merges, force-pushes, or opens PRs unattended.

Classifies owner-pasted test/CI logs into flaky, regression, lint, or deps and reports a queue-ready failure memo. Never merges or deploys.

Turns owner-pasted NVD or advisory exports into CVE alert and triage memos matched to your deps list. Never bumps versions or auto-upgrades.

Drafts changelogs and release-prep notes from owner-pasted merged PR history. Never tags or publishes releases.

Turns critical CVE, unsafe-patch, or blocked-queue signals into short owner alert memos. Notify/draft only — no autonomous merge or send.

Reads public SEC filings (8-K, 10-Q, Form 4) on an owner watchlist or pasted docs and flags material changes with source line cites. Research only — never trades or invents numbers.

Cross-checks numbers and stress-tests a market thesis against public filings or pasted docs before it ships. Refuses claims without source lines. Never trades.

Owns the brand brief and content strategy, then stages an approved brief with topics, hooks, and sources ready for drafting. Never publishes.

Turns approved briefs into platform-specific drafts using hooks and repurpose patterns. Draft only — no schedule or publish.

Independently grades drafted posts, returns top fixes, and marks approved-for-human when clean. Never publishes.

Runs final checks on approved drafts and stages a schedule checklist plus copy for the owner to paste into their own scheduler. Never auto-posts.

Runs Monday community ops from owner-pasted Circle Insights/CSV/exports: quiet members, comment-leak posts, draft check-in DMs. Never sends unattended.

Breaks a weekly GTM mission into work packets and done criteria. Refuses incomplete done marks. Does not send outreach.

Builds GTM discovery probes from owner goals: interview questions, evidence quotes, and experiments that can fail. Research only — never sends invites.

Stores short market claims with sources as memory — not essays. No outreach.

Drafts public GTM language only from approved intel and market signals. Draft only — never publishes.

Qualifies accounts against ICP using owner-pasted firmographics and notes. Stages a qualify memo only — never sends mail.

Tracks Cursor and Grok Bot weekly spend from owner-pasted invoices or usage exports and warns at 25/50/75/90% of a budget you set. Never pays, never freezes bots, never kills spend.

Rewrites owner-pasted drafts to remove AI telltale phrasing while keeping the voice and facts. Draft rewrite only — never posts or sends for you.

Finds local events from public Luma/Eventbrite/listings the owner points at and builds a Monday one-pager of worth-attending picks. RSVP only when the owner is present to confirm.

Builds a two-column audit of recurring charges versus last-touch evidence so you can decide what still earns its keep. Never auto-cancels subscriptions.

Turns business podcasts or YouTube episodes the owner points at into short no-fluff cheat sheets you can act on.

Assembles a diligence pack from owner-pasted Drive/folder notes for asset sales (Airbnb/multi-unit pattern). Never lists the property or sells alone.

Reads an insurance denial the owner pastes, surfaces the appeal deadline, and drafts an appeal letter for a human to file. Never files the appeal alone.

After you publish an X post, surfaces people and ideas worth a reply, bookmark, or follow-up draft. Never sends DMs alone.

Scouts used Teslas near a ZIP against your must-haves and returns a short matching list with a staged inquiry email for you to send. Never buys or messages sellers alone.

Audits your bot schedules for noisy polling, all-clear spam, and overlapping roles. Silent when healthy. Never unpauses what you froze.

Drafts the leasing workflow for a property — ad, terms, listing copy, prospect Q&A, and lead follow-ups — for you to review and send. Draft only; never publishes listings alone.

Inventories your bots and connectors, severity-tags access, runs a weekly security scorecard, and reviews outbound drafts to clear, redact, or escalate. Read/inventory only — no send, pay, create, or delete.

Creates post stills and teaching graphics so you never open a design tool mid-route. Delivers files for human review — never auto-posts.

Ingests host snapshot packages (ZIP, checksum, timestamp) and flags performance or security drift on a watched machine. Monitor and alert only — never remediates unattended.

Owns SwiftUI feature work and screenshot proof. Drafts code and review notes — does not merge.

Edits Espanso YAML snippets, checks trigger collisions, and scouts new triggers so you stop hunting config files. Local files only — never phones home with snippet contents.

Uses a browser WhatsApp Web session to list, read, and search chats. Sends only with the owner present to confirm. Capture-first inbox desk — no mass messaging.

Triages calendar collisions, trips, paperwork, and bills into next steps and usable drafts. Never pays and never books travel alone.

Builds a searchable, sortable local table of your X bookmarks (author, date, post, engagement), opens details, and pages more when you ask. Read-only on bookmarks.

Helps with Discogs collection and wantlist checks, marketplace price reads, and photo ID of records. Adds to cart only with your confirm — never buys alone.

Builds a profile from your CV, scores public job postings, and returns a ranked shortlist with draft cover notes. Never applies alone.

Tracks the small loops you promised on calls, email, or Slack — intros, assets to send, prospect asks — then reminds you and drafts the follow-through. Draft/remind only.

Produces daily TikTok slide images and captions on your cadence; you post. Reads analytics you paste and iterates. Never auto-posts.

Learns your grocery list and AI tool stack, then only pings matching Coles/WW/Aldi/Costco specials, quota resets, and checkout codes. Quiet when nothing matches. Never pays — you check out.

Maps plain-English laundry loads to real LG ThinQ washer/dryer cycles. Starts Remote Start only with you present to confirm. Never starts a cycle alone.

Charter guard for Link shopping: per-order and rolling-window spend ceilings, known-merchant rules, and an approval log so casual thumb-taps cannot greenlight bulk restocks. Never pays — human still approves.

Audits your scheduled routines monthly or when you add one: flags empty handoffs and reminder spam, then proposes keep or drop with evidence. Never deletes or pauses routines alone.

Installs Calibre on the bot computer and converts ebooks in chat: drop a file, name the format, get the converted book back. Local conversion only — never republishes or uploads to stores.

Cuts cinematic short drafts, then honestly reviews GitHub and skill listings and rebuilds the one that actually runs. Draft/cut only — never auto-posts.

Scans your X mentions for negative or hostile replies and proposes mutes for those profiles with a short reason. You confirm each mute — never mutes alone.

Finds remote-meal receipts in work email, matches them to Ramp expense transactions, and attaches only after you approve. Never submits expenses or pays.

Finds and compares round-trip flights, builds public Skyscanner-style search links, and watches routes for bag-included deal drops. Surfaces shortlists only — never books or pays.

Starts from invention disclosure forms, watches named GitHub repos you point at, and after your OK drafts disclosure and write-up packages for counsel. Never files patents or emails counsel alone.

After flights and hotel are locked, maps the trip day by day. Asks OK before peeking at booking email. Drafts itinerary only — never books or cancels.

Where to stay before you book: neighborhoods, stay type, day skeleton, food landmarks, local secrets, and a soft etiquette card. Drafts only — you send the bookings.

Benefits navigation for employer/plan paperwork — surfaces steps, forms, and decision checklists. You submit. Never files claims or talks to insurers as you.

Helps prepare and navigate App Store Review — checklists, reply drafts, and rejection triage. You submit in App Store Connect. Drafts only.

Runs Midkernel Scan from chat in a sandbox, then returns the report and full log. Midkernel is a third-party product metered by credits. Report/log only.

Flags large insider buys from public filings/news (e.g. big Form 4 purchases) and surfaces them for you. Read-only alerts — never trades. Not financial advice.

Uses what already landed, proposes 5 sourced ideas, then draft → critique → rewrite → your OK. LinkedIn drafts wait for your go-ahead. Never auto-posts or auto-engages.

Daily X review and topic bank. Drafts posts only after you approve. Hook checklist baked in. No spammy auto-engage — never likes, replies, or DMs unattended.

Turns non-urgent mail the owner forwards or pastes into a curated personal newspaper with sections, ads skipped, and asks highlighted. Does not send mail. Does not open a mailbox.

Drafts contract redlines on a document the owner pastes, with a short rationale per change. Never sends to counsel or the counterparty. Not legal advice.

Attacks the verified lead and lists what would falsify it. Does not trade. Does not post.

Matches owner-pasted invoices (PDF/CSV/mailbox exports) to payments, flags mismatches, and drafts chase notes. Does not send mail, move money, or touch Stripe. Owner must be present to send any chase.

Drafts RFP and security-questionnaire answers from docs the owner pastes. Owner signs off on every answer. Never submits the questionnaire.

Sources inventory, stages carts, and grades plays with real-fee margins. Never checks out. Never pays. Human keeps the buy call.

Rebuilds payouts, fees, receipts, and a monthly P&L from owner-pasted CSVs. Never moves money. Never pays.

Triages daily must-move storefront issues and stages the next action. Does not checkout. Does not message buyers.

Maps prior cycles, named actors, and what happened last time this narrative ran. Does not trade. Does not post.

Drives a product-return flow end-to-end with the owner present: checklist, portal steps, draft messages, and tracking notes. Never files the return alone. Never chats with the merchant unattended.

Turns owner-pasted win/loss call notes into a structured memo (why won/lost, competitors, next plays). Draft only. Does not update CRM or email anyone.

Runs a 7-step product-launch VIDEO playbook: brief, script, shot list, on-screen text, CTA, cut list, and thumbnail copy. Does not publish to YouTube, TikTok, or Instagram. Does not buy ads.

Digests new model releases from public web and X the owner cares about: API cards, pricing, context window, and license. Not a trading bot. Does not buy, sell, or tip a ticker.

Writes the final research memo for the owner from the pipeline files. Does not tweet. Does not trade. Waits for human approval.

Reads overnight Slack the owner scoped and writes a morning audio-brief SCRIPT in chat (text for the owner to read or record). Does not post to Slack. Does not generate or upload an audio file unless the owner asks locally.

Extracts who promised what by when from meetings and Slack the owner points at, and keeps an open-promise list over time. Recaps in chat. Never pings the promise-maker unattended.

Watches the seller tool and storefront for bugs, then recaps a ping for Build in this chat. Does not send Slack unattended. Does not merge.

Turns public Reddit pain threads into a microSaaS research brief. Does not create Reddit accounts, auto-comment, vote, or ship product.

Clusters a noisy Slack channel the owner names into themes, stated owners, and open questions. Does not reorganize the channel, delete messages, or post the list unless the owner is present and asks.

Diffs public competitor pricing pages from a URL list the owner maintains into a price-row change sheet. Narrower than a full competitive brief. Does not scrape behind login. Does not email.

Inventories existing bots, skills, and routines, backs up notes, and proposes a move into this org system. Does not delete agents.

ICP-screens webinar or event registrants from a CSV the owner pastes, using CSV fields by default. Public LinkedIn or company-site checks only if the owner asks. Does not email registrants.

Engineering on the seller tool only. Escalates real code to a cloud agent and babysits the PR. Does not merge. Does not checkout.

The only seat in the feed. Scans public crypto narrative chatter and stages leads for the verifier. Does not trade. Does not tweet.

Watches a Gmail label the owner names (or a paste/export) so SMB quotes and payments do not slip: flags quotes awaiting reply, unpaid invoices, and unanswered threads. Drafts follow-ups. Sends only while the owner is present. Never sends unattended.

Verifies Radar leads against public sources and marks each claim sourced, weak, or false. Does not trade. Does not post.

Owns one active project the owner names. Tracks blockers, borrows shared specialists, and recaps in chat. Does not spawn extra bots. Does not send unattended.

Single router on top of Scout, Build, Ops, Ledger, and Sentry. Hands the owner a recap and the decisions only they can make. Does not checkout, pay, or merge.

Shared specialists that project leads borrow for research, writing, or numbers. Does one task and returns a file the project lead can inspect. Does not send.

Coach map literacy from owner-pasted map images, legends, or verbal map descriptions: scale, orientation, grid refs, and short route plans. Asks for the student's attempt first; never invents places or distances.

Helps students read climate graphs and tables from owner-pasted numbers, CSV snippets, or chart descriptions: axes, seasonality, anomalies. Tutor hints first; never invents climate data.

Explains physical geography processes and landforms from owner-pasted syllabus questions or textbook excerpts. Tutor mode with one-hint steps; never invents site-specific measurements.

Scaffolds human-geography case answers from owner-pasted case packs, articles, or worksheet facts. Demands student attempt; never invents population or economy numbers.

Builds and critiques cause–effect timelines from owner-pasted event lists or syllabus chronologies. Tutor hints first; never invents historical facts or dates.

Coaches OPVL / origin–purpose–value–limitation critiques on owner-pasted primary or secondary sources. Asks for student attempt; never invents provenance.

Helps students compare periods, regimes, or turning points using owner-pasted briefings or syllabus notes. Structured compare tables; no invented facts.

Scaffolds history essays from owner-pasted prompts and evidence packs: thesis options, paragraph plans, citation checks. Will not write a full graded essay for submission without disclosure.

Tutor for algebra manipulations from owner-pasted problems: ask for attempt, one hint, then check steps. Never silently finishes the exam set.

Reads and explains functions and graphs from owner-pasted equations, tables, or graph descriptions. Tutor hints; no invented axis scales.

Turns owner-pasted word problems into quantities, relations, and equations. Asks for student decode attempt first; then tutors the algebra without taking the exam silently.

Diagnoses common algebra misconceptions from owner-pasted student work, then drills a corrected mini-pattern. Tutor honesty: attempt first; no silent exam completion.

Coaches geometry proofs from owner-pasted givens and diagrams-as-text. Hint-first; never invents unmarked lengths or angles.

Computes and tutors length, angle, and area problems from owner-pasted figures and measures. Hint-first; no invented side lengths.

Tutors geometric transformations from owner-pasted coordinates or diagram specs. Hint-first; never invents preimage points.

Descriptive stats coach from owner-pasted datasets or summary tables: center, spread, shape, outliers. Tutor hints; never invents n or values.

Helps students read statistical graphs from owner-pasted chart text, screenshots-as-description, or table+graph pairs. No invented axis ticks.

Classroom intro to statistical inference reasoning from owner-pasted summaries or prompts: sampling talk, confidence intuition, and claim checks. Tutor hints first; never invents sample size, SE, or p-values.

Portuguese grammar coach for owner-pasted exercises: agreement, verb forms, punctuation. Asks for student attempt; feedback and corrections — not silent full-worksheet answers.

Guides interpretation of Portuguese literary or informational texts from owner-pasted passages and questions. Tutor hints; no invented quotes.

Applies owner-pasted ENEM/school rubrics to Portuguese essay drafts: criterion feedback and revision targets. Will not ghostwrite a full graded essay for undisclosed submission.

Line-level Portuguese rewrite coach for clarity, cohesion, and register on owner-pasted drafts. Shows options; keeps student authorship with disclosure norms.

English grammar coach for owner-pasted exercises: tense, agreement, articles, punctuation. Attempt-first feedback; no silent full-worksheet completion.

Supports English reading comprehension from owner-pasted passages and questions. Hint-first; evidence from the paste only.

Scaffolds English paragraphs and essays from owner-pasted prompts/drafts: thesis, structure, evidence use. Will not ghostwrite a full graded essay for undisclosed submission.

Runs English listening-style practice from owner-pasted transcripts or captions only. Does not invent audio or pretend to hear sound.

Culls travel and family photo batches into a curated folder with a consistent, light touch-up recipe. Never deletes or overwrites originals; asks before removing people or objects. You confirm destructive edits.

Builds the cheapest review-ready Instacart cart from your list: learns store prefs, verifies live prices, stages the cart. You check out — never pays.

Proactive LinkedIn digest on a schedule: new DMs, invites, profile views, matching posts, and job searches — surfaces digests in chat. Read/watch + digest only.

Runs affiliate recruitment research: topic and competitor searches, scoring, shortlist, and contact outreach drafts. Research + drafts only — you send.

Answer Engine Optimization (AEO): turns AI-visibility data into article briefs and drafts aimed at showing up in AI answers — brief → draft → editorial, then stops for you to publish. Never posts to CMS unattended.

Reviews Gong call recordings (paste/upload) and returns short, actionable post-call coaching bullets for managers and enablement — what to fix before the next call. Draft coaching only.

Scores sales calls from a pasted transcript or uploaded recording and tells you what to fix before the next one. Works without Gong. Coaching notes only — never emails the rep or updates CRM unattended.

Overnight freelance deal desk: scores inbound leads against your rate and stack, drafts scoped proposals and calendar holds, and wakes you on client reply or <12h deadlines. Drafts only — never sends email or proposals unattended.

Native Apple/Mac development on your connected Mac — Xcode builds, simulators, Swift, and remote verify from your phone. Never ships to App Store Connect or TestFlight unattended.

Turns uploaded invoices and financial statements into structured JSON and Excel. Native-text PDFs extract on-box; scans are classified then extracted. Extract/structure only — never files taxes or emails banks.

Takes an offer from nothing to a review-ready Whop store page in one run: product, pricing tiers, checkout links — then waits for you to publish. Never spends ad budget.

Monday P&L for Whop businesses: revenue, MRR, churn, refunds, ad spend, week-over-week with every dip explained. Strictly read-only — never moves money or changes plans.

Runs Whop paid ads end-to-end as drafts: creative, launch plan, daily triage, morning report — starts small and cuts losers. Asks before raising budget. Never spends without your approve.

Catches Whop cancellations before they land, drafts the cheapest real save (uncancel/pause), and reports why people left. Offers saves for your confirm — never charges or refunds unattended.

Runs clipping/UGC bounties: ungaming briefs, cost-before-escrow, submission checks, owner-ready payout recommendations. Shows cost before escrow. Never pays creators without your approve.

Runs the Whop affiliate program: commission terms drafts, performer ranking, cut bad actors, arm good ones with assets. Drafts terms/cuts for your approve. Distinct from affiliate-scout (recruit research).

From calendar or a guest-list screenshot, builds a face/name cheat sheet and detailed guest research report for dinners/events. Clay via paste or MCP-when-user-present only. Research report only — never messages guests.

Turns one seed company into lookalike accounts and verified decision-makers for outbound. Research + shortlist only — you send. Wedge vs any live lookalike/lead bots: research-only living list, never outbound send.

Watches rival companies for funding, hiring, news, and ads so GTM teams catch momentum early. Public-source watch + digest. Never impersonates or messages rivals.

Tracks SERP and AI/LLM visibility so content teams know what to publish next. Visibility digest + brief only. Never publishes content unattended.

Implements approved motion specs with official GreenSock GSAP skills — timelines, ScrollTrigger, framework cleanup, reduced-motion path. Code/implement only after approved spec. Never ships to prod unattended.

Finds money that should be yours but is missed, wasted, owed, or unclaimed — then surfaces a recovery list from public sources / owner-present evidence. Discovery + list only. Never files claims or moves money. No financial advice or fraud.

Helps challenge unfair parking tickets, fines, bank/card fees, deposits, travel refunds, and tax penalties with draft dispute packs. Drafts challenges only — never files, pays, or submits disputes unattended.

Runs Alibaba factory sourcing end-to-end from SOPs: onboard model/tracker, drive UI with owner present, file cards — never pays or confirms payment without your yes.

Hunts forgotten SaaS in Gmail (owner paste/export), scores a Zombie Index, opens an HTML autopsy with optional cancel paths. Read-only by default — never cancels subscriptions unattended.

Farms action items from meeting notes and Slack into the tracker — dedupes, triages, files clean draft issues with sources. Creates draft issues only where policy allows; never pings people unattended.

Paste a job req → ranked candidate shortlist + personalized email copy via research/ops the owner provides. Nothing sends until wording is approved. Draft outreach only.

Maps who currently works at a named company by function, seniority, and location (CSV+HTML census). Org map, not candidate search. Research/export only — never messages people.

Vets inbound LinkedIn invitations with a Peer-first preset; drafts accept/ignore/personalize recommendations only after you confirm. Distinct from linkedin-watch-desk digest. DRAFT invites/replies only — never send or accept unattended. Never stores passwords.

Turns podcast episodes into social quote graphics and carousels. Paste episode, pick nuggets, drop template ID (ContentDrips-style via owner tools). Creates drafts/assets only — never publishes unattended.

Finds hot posts in your niches and drafts short replies with a weekly spend/activity cap so you grow on X without spray-commenting. You still send every reply. Draft replies only — never posts or replies unattended.

Learns any X handle’s writing style from their posts, then drafts tweets, edits, and long-form in that voice. Drafts only — never publishes unattended.

Plans trips and drafts bookings for your approval — itinerary, options, booking drafts. Never pays or confirms bookings without your yes. Distinct from flight-deal-scout (deal watch).

Child app-usage safety monitor (YouTube, Family Link, Screen Time style drops) — what kids open and how often, not a content wiretap. Daily usage report. Owner-scoped accounts only. Never changes device settings unattended.

Imports WeChat/Alipay bill CSV or screenshots into a personal ledger: parse, dedupe, categorize, preview before write. Never touches payment passwords or proxy-login. Write only after preview confirm.

ADHD / executive-dysfunction helper: short questions that get you moving on a stuck task one step at a time (or an honest out like half-ass / skip). Coaching prompts only. Never sends messages or spends money unattended.

Japanese business-writing checker for boss/approver-facing copy: grammar, orthography, keigo, readability — keeps meaning, no new drafting or sending. Edit/check only.

Finds used Teslas in the US by ZIP + model (3/Y/X/S) and drafts matching listing digests/emails — warm, plain, no hype. Listing digests / email drafts only. Never buys or messages sellers unattended. No financial advice or fraud. Public listings / owner-present only.

Psychology literature study desk: onboards your field and journal cadence, finds citable references, writes dated bibliographic surveys and article notes, asks studied/discarded/later checks, and keeps a running study chronology.

Turns a city + business type into a clean lead CSV via a local open-source Maps scraper the owner runs (Docker on localhost). Stages outreach drafts only — never emails or SMS unattended. Warns on rate limits, Google ToS, and PII law before big jobs.

Finds job openings that match the queries you give it across public LinkedIn, X, and company career pages and returns a short ranked list with links. Search and report only — never applies or messages recruiters.

Upload a resume and get matching roles with fit scores, tailored materials, interview packs, and a follow-up tracker. Applies only with you present — never auto-submits applications.

Sources unusual early-stage startups for angel investors — under $1M raised, strong teams, chosen geographies — from public sources, and delivers a shortlist table whenever you ask. Research only — never contacts founders or moves money.

Prospecting desk for agencies selling to Shopify stores: finds public-signal leads (new launches, hiring, ad activity) and drafts founder-voice outreach. Leaves drafts for your approval — never sends.

Turns a YouTube lecture you link into a dense, exam-ready cheat sheet: cited formulas with timestamps, key diagrams described, and a math check before delivery. Document generation only — never posts or shares the sheet.

Cuts AI/ML papers into claim, use case, and math. Reads the PDF itself — not just the abstract — and returns a one-page Paper Card for practitioners. Research cards only — never publishes or emails authors.

Reputation desk for local and online businesses: when you ask, checks the Google, Yelp, Trustpilot, Facebook, and Amazon review pages you name for new and unreplied reviews and drafts on-brand replies. Draft replies only — never posts.

Triages customer support for Linear: retitles in-app feedback, turns Slack and HubSpot reports into factual ticket drafts ready for Linear that you file yourself, and adds a short diagnosis. No Linear plugin is bundled. Never contacts customers.

Client-board helper for freelancers and agencies: reconciles a Trello board export with your tasks sheet, drafts a status digest per client, and turns inbox replies into tracked updates. No outbound client mail without your approval.

Tracks ocean bills of lading for importers from carrier tracking pages you list and writes an exception-first digest on request: transit days, ETA slips, demurrage risk. Status digests only — never books freight or pays carriers.

Works on the scanned files you share with it: reads each one, proposes a descriptive name and destination folder, and flags duplicates by hash once the files themselves are shared, so originals can be recycled with confidence. Never deletes or shares without your confirm.

Finds people who completed your order form but did not pay, and drafts a personalized recovery email with free proof of work. Draft emails only — never sends recovery mail or charges cards.

Family mission control for school chaos: reads school emails and portal exports you forward, checks which guardian already signed up for what, and builds one family calendar digest. Never emails teachers or signs forms.

Keeps a multi-bot roster documented from a Notion-style table you paste — routines, channels, rules — so your fleet can be rebuilt 1:1 elsewhere. Roster sync drafts only — never pushes live bot configs.

Your home-hunt specialist: locks city, budget, beds/baths, commute limits, and must-haves, then scans public listings and returns a shortlist you can act on. Search and shortlist only — never messages landlords or applies.

Fridge and pantry photos become real recipes, a grocery list on your budget, and new ideas from what you already have. Recipes and lists only — never places grocery orders.

Builds a dated HOA violation-response file and draft letter from your notice and evidence: timeline, rule cites, photos list. Not legal advice. You keep the send button — never emails the board or pays fines.

Turns a messy memory into a dated, sourced witness-statement draft with a timeline, for police, a lawyer, or your own records. Not a lawyer. Draft only — never files or contacts parties.

For homeowners holding mismatched contractor PDFs: builds an apples-to-apples grid of line items, allowances, exclusions, timeline, and payment terms, with blanks flagged. Compare cards only — never awards jobs or pays.

For renters before signing: a plain-English gotcha card on fees, entry, pets, renewals, and deposits, each tied to a quoted clause; unknowns flagged. Review card only — never signs or sends to the landlord.

Plug-and-play job-hunt factory: for each role, research → human-voice materials → fact and voice gates → ship or kill, then apply with you present. Never auto-submits.

Drop a GitHub repo or folder and get a two-sided FAF stack flip card. A .faf file is a small YAML project-context file (Foundational AI-context Format) that describes a repo's stack and purpose. Facts from the .faf file only — no invented scores. Card generation only — never opens PRs or merges.

Paste a diff or snippet for merge-ready review: returns only findings above 80% confidence plus an APPROVE / WARNING / BLOCK verdict — no nit spam, no vague approvals. Review comments only — never merges.

Personal digital-security guide: inventories your online accounts from what you share and walks you through locking each down on official first-party settings pages, one step at a time. Never asks for or stores passwords.

For people who want their info off people-search and data-broker sites: finds listings about you from public search and prepares official opt-out requests for you to submit. Consent-first — never submits or pays unattended.

Home upkeep from photos: learns your house, builds a filter/yard/maintenance calendar, and proposes purchases you approve. Never buys anything without your explicit approval.

Local housing beat: reads the city and county agendas and news you name and writes a short note on items that actually moved. Digest only — never contacts officials.

A backup parent for chores: takes a chore from you and follows up with your kid only in a chat channel you connect, until it's done with proof, then reports back. With no channel connected, it drafts the reminders for you to send. Never contacts anyone else.

Finds and ranks scholarships for your student — from big state awards to small niche ones — and drafts application packs. Never submits applications or pays fees.

A school assistant: reads your learning-portal export or screenshots and writes a digest of new grades, upcoming work and tests, and truly overdue items. Read-only — never submits work or messages teachers.

Blunt pre-owned deal hunter across eBay, Facebook Marketplace, Mercari, OfferUp, Craigslist, and Reverb: finds deals for your wish list and filters scams. Never buys or messages sellers.

Finds accelerators and pre-seed VCs for early-stage founders, matched to location and sector, as a shortlist table you can track. Never contacts investors or submits applications.

Finds people publicly asking for your product or service near you — you set the category and distance — and brings matching posts with links. Never replies or DMs for you.

Local-first agricultural sourcing: finds, qualifies, and ranks real growers and suppliers for an ingredient, closest high-quality source first. Never places orders or contacts suppliers.

Trust auditor for Grok Bot agents, skills, routines, and shareable templates: flags silent phone-home, secret harvesting, and undeclared creator telemetry before you add them. Audit only — never installs or deletes.

Type an address or neighborhood and get one digest of what is changing around it: building permits, service requests, meeting agendas, business and property registries, and site plans from public city and county sources, grouped by distance with links. Digest only — never files complaints or contacts officials.

Takes a day-to-day errand — an appointment, a service booking, a reorder, a cancellation — compares options, and hands you one recommended proposal. You book or pay yourself; then it checks the confirmation you share, drafts the calendar entry, and keeps a one-line proof packet. Never books, buys, sends, or cancels on its own.

Runs the ops side of awards banquets, galas, and festivals from what you share: vendor and sponsor commitments, missing paperwork (COIs, W-9s, contracts), volunteer roles, a day-of run sheet, and polite outreach drafts. You send every email — never commits spend or messages vendors on its own.

For people representing themselves in Texas district, appellate, or Supreme Court matters: keeps each case separate with its docket entries, computed deadlines to verify, filing to-dos, and a document shelf from what you share. Not legal advice and not a lawyer — tracks and reminds; never files anything.

A one-page tax-year discussion brief for households without a W-2, roughly ages 55–80: MAGI vs IRMAA tiers, RMD timing, Social Security taxability, estimated payments, and questions for your preparer — built from numbers you share. Not a preparer and not tax advice. Never files, pays, converts, or trades.

Finds threads with your VIPs where they are waiting on you and you went quiet — from the Slack or email threads you paste in — ranks them by who and how long, and drafts a paste-ready follow-up in chat. It has no Slack or email access and never sends anything — you paste and send.

A safe sandbox for one long agent prompt at a time: spell-check and format, resolve ambiguous instructions, cut tokens from skill and routine bodies, and return reusable paste-ready text with a change log. It never runs the prompt or sends it anywhere mid-draft.

For nuclear-beat journalists and analysts: searches the NRC's public ADAMS document library for new filings by docket, plant, or topic you pick and highlights the ones worth talking about, plus the US fleet's operating share from NRC power-reactor status reports. Digest only — never files with the NRC or contacts regulators.

Shops everyday goods for you when you ask: finds and compares listings on Amazon and other retailers' public pages, using the preferences and past orders you share, then hands you a cart-ready pick with price and options. You check out yourself — never pays, and never asks for passwords or card numbers in chat.

A local Wi-Fi telemetry analyst for your own space: reads Wi-Fi scans you paste (or export from your own device), charts channels, signal, and congestion, and labels every claim with how sure it is. It never invents occupancy or distance from signal strength and never joins or probes networks.

Owns roofing lead intake from the inquiries you share: qualifies each one, drafts SMS and email follow-ups, and preps a quick-quote handoff packet for the estimator — with public business-info (NAP) consistency and founder-privacy rules built in. Drafts only — never texts, emails, or quotes on its own.

A field sales OS for reps who cover an installed base of customer sites: builds the day plan from your accounts and visit history, maps the installed base, walks the rep through a Free Site Audit worksheet that also trains them, and rolls up a light leadership view. Plans and worksheets only — never sends audits to customers or commits quotes.

Org briefings at the right altitude from the same signal you share: a standup note, a team update, a department review, a board summary, and a shareholder letter section — same facts, different density. Briefings only — never posts digests to Slack or email or changes org records.

Turns a bottleneck diagnosis or a workshop decision into an exact change plan for your org model — what to add, rename, reassign, or retire, with owners and acceptance checks — and hands it to you to review and apply yourself. Built for teams that keep an org ontology (projects, people, processes); the bot never applies changes to any system.

Multi-day RV trip planner for full-timers: builds a chat trip board around your rig length, hookups, drive-day limits, and pets, checks campground availability on public pages, and tees up each booking with the link and details for you to review and pay. Never pays or confirms a booking.

Turns your Instagram Saved into a personal media wiki: reads the data export you download from Instagram, transcribes the clips you share, tags and facets each item, and builds a cited wiki you can search. Your media stays in your files. Never posts to Instagram, signs in as you, or deletes anything.

For restaurants on delivery apps: finds canceled orders and error charges in the merchant-portal exports you share and stages a reimbursement packet for each — order ID, amount, reason, evidence. You submit the claims yourself; it never files reimbursements or disputes.

A team data analyst: ask a question and get a report or chart from a confirmed org schema and the event logs or query results you share. Read-only — it never changes the data model or writes back to any store.

An ETL steward for after your org schema is confirmed: designs how event logs from your sources get extracted, canonized, and loaded into your datastore, and writes the mapping and load scripts for you to run against staging first. It never deletes sources or ships production loads.

A coding coordinator for developers who use Cursor cloud agents: splits work into one job per agent, writes each launch brief with fixed facts (repo, constraints, acceptance tests), then reviews the agents' results against those checks and reports in plain words. You launch the agents; it never merges PRs or pushes to main.

Finds like-minded people on X worth engaging with — an even mix of fast-growing, regularly active, and large established accounts in your niche — and recommends specific accounts and posts, retuned by your feedback. Recommendations only: it does not draft replies and never posts, replies, follows, or likes as you.

Cleans up your X followers and following: from the lists you export or paste, ranks likely bots, spam, dormant, and off-topic accounts into block and unfollow lists with profile links, sized to the per-run cap you set. Lists only — you block or unfollow; it never acts on your account.

A phone-first conflict coach: helps you name your side clearly, separates facts from feelings and needs, then drafts a calmer message to the other person in your voice. One relationship slot per person, kept separate. It never reads the other person's chats and never sends — every message is yours to send.

For domain investors who use DomainSnip: reads the keyword-alert emails or lists you paste and flags notable new and ending-soon .com matches — price, length, marketplace, end time — with a short why. Also helps tune your alert keywords. Never buys, bids, registers, or transfers domains.

Designs a four-bot newsroom around your beats and writing voice — two research bots, an editor desk, and a fact-checker, plus a group room — as ready-to-create bot specs. Then runs the flow you feed it: briefs → draft → fact check → social cut and a teleprompter-ready script. You create the bots and publish; it never posts or goes on air.

For YouTube creators posting ride and FSD dashcam footage: cleans up titles, descriptions, and playlists, builds highlight-reel cut lists with timestamps (and renders draft reels from clips you share), and drafts thumbnail concepts. Publishes nothing — you upload when you are ready.

A public-records style research desk for people who want to dig themselves into public US equity settlement data — SEC fails-to-deliver files, public swap repositories, and short-interest reports — to find structural patterns worth a closer look. Structural smells only: no fraud or intent claims. Not investment advice; never trades.

Audits your own X history — posts, replies, quotes, and reposts from the archive you download or paste — for reputational risk, and sorts findings into clear, ambiguous, and screenshot-risk with links. You decide what to delete or edit and do it yourself; it never deletes, edits, or posts.

For people running TypeSafe Jev browser automation on a Chrome they can watch, with their own Jev harness and provider key: writes run goals with budgets and stop gates, and reads the step traces you paste to find failures and when to fall back to screenshot computer-use. It does not drive a browser itself and never approves payments, logins, or irreversible actions.

A mid-to-long-term stock research desk: finds and scores quality companies for a roughly five-year horizon — beaten-down names or ones moving into strength — on leadership, business quality, financial strength, and the contrarian narrative, with sources. Research only, not investment advice. Never places trades or broker orders.

A private, gentle desk for anyone settling a loved one's affairs: keeps the estate file, the list of institutions to notify, and an expense and inventory log, and drafts the letters you send. Not legal, tax, or financial advice — for probate questions it points you to the court's own pages or an attorney. Never sends, pays, or files anything.

Coordinate care for someone you look after: a care profile and running log, visit-prep cards for appointments, med cards quoted exactly from the label you share, shift handoffs between helpers, and post-visit follow-ups. Not medical advice — it never diagnoses, never changes or suggests doses, and points you to the clinician or pharmacist. Never contacts anyone.

Your private moving desk: a move plan with the critical path, mover quotes side by side with red flags, an address-change and utilities tracker, a room-by-room box log that finds anything, a move-day pack, and a settling-in close. Never invents dates, prices, or local rules; never books, pays, or contacts anyone — you do.

For parents of a child who guesses at words while reading: short lesson cards and practice exercises that slow down guessing and build decoding accuracy, plus mastery checks you run together. Reading practice only — it never diagnoses, screens for, or suggests any condition; for an evaluation it points you to a qualified professional or the child's school.

A Spanish speaking coach built for ADHD and dyslexic learners: short speaking rounds (voice where your app supports it), tap-to-answer choices, no spelling penalties, and a quick log of what stuck. Never shares recordings or progress with anyone.

A pretend family bank: keeps a private ledger per person, pays exact compound interest at the rate you set, and makes friendly statements kids can read. No real money ever moves, and it never links bank accounts.

Makes a children's picture book from start to finish: five easy questions, a story, full-bleed square art, a labeled page preview, and a print-ready PDF. For parents, teachers, and beginners. Book drafts only — never orders prints, uploads, or publishes; you decide where it goes.

Choose a life arc and answer one question at a time — typed or spoken — and it turns the answers into a printable memoir manuscript with a cover. Built for family memory books, including diaspora families. Interview questions and drafts only; never emails relatives or publishes.

Helps parents get a child's Common App complete and polished: activities and honors wording, essay feedback (the student writes), a smarter college list including sports and ROTC paths, a deadline tracker, and side-by-side financial-aid offer comparisons. Never submits applications or emails admissions.

Presentation coach for any talk, live or remote, any size: brainstorm and draft the outline, slides, and speaker notes in chat, then get dry-run feedback on your deck and a practice take you send. It does not listen live and never presents, posts, or emails attendees.

A plain-English IT helpdesk and tutor for beginners: when your computer is stuck it works through it with you from your own screenshots, one click at a time, and offers short Windows-basics drills when you want to learn. You do every click — no remote access. It only names buttons it can see, never asks for passwords, never buys anything, and never suggests a change you can't undo without your okay.

Your DIY sidekick: send a photo and get the likely cause, the part to buy, and step-by-step instructions, and it remembers where you left off on each home project. Great for first-timers. Calls in a licensed pro for gas, main electrical, or structural work. Shopping lists only — never orders parts or books contractors.

A step-by-step guide to selling your own used car privately: a price range from market comps, Autotrader and Facebook Marketplace listing drafts, safe meetup spots, and a fill-in bill of sale. Title and registration transfer steps are pointers to your state's official DMV or registry page, which you follow — not legal advice. Never posts listings, messages buyers, signs, or files anything.

A genealogy research desk built on analyst tradecraft: searches free public historical records first, rates every source, tests competing hypotheses, and names the weakest link in your family line — from public sources and what you share. Deceased ancestors only: it never searches or profiles living people. Warns that research is token-heavy and lets you pick the run size. Never pays for sites or requests records for you.

Helps make an inbox sane: writes Gmail filter rules that archive promo and junk (never delete) for you to add, and turns mail you paste or forward into a short bills-and-events note kept as a draft. It has no access to your mailbox — you apply the filters and send the note if you want.

For small service businesses (movers, HVAC, plumbers, landscapers, cleaners): drafts after-hours call scripts and voicemail greetings, website-chat reply templates, and triages the messages or voicemail transcripts you paste — who should handle each and which are same-day or next-day rushes. It does not answer phones or chat by itself: there is no telephony or chat integration, so you or your answering tool use what it writes.

Back office for small field-service crews: paste what your crew texted you and it keeps the job board, flags down equipment and unpaid invoices, and gives you one short brief whenever you ask. Never texts customers, sends invoices, or charges anyone.

Residential maintenance ops desk for property managers: intake → P1–P4 triage → dispatch briefs → resident status-text drafts → closeout photo QA and unit notes, with hard stops on spend, unit access, and life safety. Drafts tickets and texts only; never texts residents, approves spend, or closes work orders.

Phone-first construction-manager desk for US civil work, owner-side or contractor-side: site-visit cards, photo → description → issue → RFI draft packs, spec and public code checks, and OAC meeting prep. Drafts stay in chat for you to finish; never emails contractors or closes RFIs.

Finds grants your small business can actually win, screens each for fit against eligibility, and drafts applications for your approval, with a short board digest whenever you ask. Never submits applications or emails funders.

For web designers: finds local businesses with missing, dated, or broken websites, checks each site's problems on desktop and phone views, and drafts friendly cold emails you review and send yourself. Never emails business owners or buys domains.

Runs one Google Business Profile for a local business: drafts posts and review replies, audits the listing, works through Needs Attention items toward full completeness, and keeps a content calendar — from the listing details and reviews you paste. Nothing publishes: you post. Not an SEO agency, CRM, or multi-location tool.

Handles a YouTube channel's post-upload work from the video or transcript you share: transcript and captions file, chapter timestamps, title options, description, and thumbnail concepts, all paste-ready for Studio. Never publishes, unlists, or edits live videos — you paste.

Alt text specialist: drop in one image or a set and get ready-to-paste alt text for blind and low-vision readers, up to 1000 characters each, with a clear warning up front when content is sensitive, violent, or gory.

DMARC specialist for email authentication and deliverability: reviews aggregate reports you share, separates real failures from forwarding and security-gateway noise, and guides SPF, DKIM, and DMARC policy from monitor to quarantine to reject. Report and DNS-record drafts only — you publish records.

Defensive pre-ship gate for code bots wrote — in your own repos and apps: checks for stale dependencies, fake or typo-squatted packages, demo-grade auth, and missing controls, then makes a ship or don't-ship call with fixes. Your own code only; never tests third-party targets. Review report only — never merges, revokes secrets, or changes prod.

An app and game QA tester that plays your web app, or works from the mobile screen recordings and screenshots you share, like a seven-year-old first-timer and like an adult QA reviewer, then writes clear, reproducible bug reports tagged by severity. Bug reports only — never ships builds, closes tickets, or messages users.

For anyone who manages a WordPress site: proposes each change as exact steps or WP-CLI scripts for your staging site, gives you checks to verify it, and only then a go-live checklist. It never publishes or changes the site itself — you run every step.

Reviews your Cloudflare setup — DNS, zones, SSL, WAF, Access, Workers, Pages, R2, D1, KV, Queues — from `cf` CLI or dashboard output you paste, then proposes changes as a diff with the exact commands. It never changes anything itself: you run each command, DNS and firewall changes included.

For Home Assistant owners: from the states, logs, and configs you paste it finds dead devices, offline cameras, failed updates, and broken automations, and writes YAML automations and fixes for you to apply. It has no connection to your Home Assistant, controls no devices, and never touches locks, alarms, or safety alerts.

Keeps a short Watching list of 3–5 public pages and the specific thing you care about on each, and when you ask, checks whether that thing actually changed since the last check. Quiet unless something moved. Never submits forms or pays for monitors.

Authentic interview simulator: give it the job and your resume; it researches the role, runs a realistic mock interview (voice where your app supports it), holds scoring until the end, delivers a blunt in-character hire/no-hire, then coaches you with concrete fixes. Never contacts employers or applies for jobs.

Paste a pay stub and get a line-by-line mismatch check against the rates and hours you provide — it never invents rates — plus a polite correction request to payroll that you send yourself. Not legal or financial advice: it points to your state labor agency's official pages or an employment attorney. Never contacts your employer or payroll and never files claims.

Paste a debt-collection notice and get a response clock built from the dates on the notice, a phone script that asks for written validation and doesn't promise payment, dispute and validation-request drafts, and an evidence checklist. Not legal or financial advice — it points to official consumer-protection pages or a licensed attorney. You send everything yourself; it never calls collectors or mails anything.

For US insurance agency staff — CSRs, account managers, producers: drafts certificates of insurance (ACORD 25, 27, 28) and new or renewal applications (ACORD 125, 126, 127, 140, plus schedules of values) from the policies, dec pages, binders, and intake notes you share, and checks filled forms against the policy. Drafts only — a licensed agent verifies every form. Not legal or insurance advice; not affiliated with ACORD. Never submits to carriers or emails certificate holders.

A paperwork and records keeper for small businesses: preps license, DBA, and state compliance filing drafts, tracks deadlines, and keeps records organized. For trademarks it only points you to the official USPTO pages — a clearance search and opinion is a job for an attorney. Not legal advice; it cites official agency pages. You review, sign, pay, and file; it never submits filings or emails agencies.

Your private desk for a child's IEP, 504, or school support plan: a child file, request drafts, and meeting packs built from the school's own documents. Timelines are quoted from school notices, and services and progress are logged against each goal's exact wording. Not legal, medical, or diagnostic advice — it points to the school's procedural-safeguards notice or an advocate. Never sends, signs, or files.

For one disrupted flight — delayed, canceled, or a missed connection: organizes the facts and receipts and drafts your own claim letter, citing only the airline policy and passenger-rights text you provide. Flight disruptions only, not general fee or fine disputes. Not legal advice. Never rebooks, contacts the airline, or files a claim — you send it.