Autopilot for your engineering backlog

Coding agents that never stop.
Fixes that never ship unreviewed.

Every run uses all your subscriptions, hands off the moment one hits a limit, and keeps going in the cloud. Errors become reviewed pull requests while you sleep. Nothing merges itself.

Runs remotely, 24/7 · $2 starter credit · No card

Sentry

PostHog

Grafana

Datadog

Linear

GitHub

BOT-42 · Todo · Autopilot · Issue intake

Checkout total wrong after quantity change

Sentry09:41

New error in acme/storefront · release 4.12.0 · 312 users in 14 minutes. TypeError: cannot read 'workspaceId' — checkout/summary.ts:88

Issue intake · autopilot09:41

Filed BOT-42 · High. Linked to release 4.12.0. Handed to the Fix & review workflow.

Operator09:42

Sentry reported a new checkout error hitting 312 users. It started with release 4.12.0. Fix & review is running on your Claude subscription; the pull request will land here for your approval.

Smart routing09:44

Claude alex@ hit its 5-hour limit at 33% of the run. Continued on your second Claude account, ops@ (62% left). Same branch, same context, 0s lost. Nothing to log into.

AAlex09:48

Approve and merge.

Operator09:49

Merged and deployed as 4.12.1. Regression watch has the error for 24 hours; Sentry shows 0 new events since the deploy. BOT-42 is done.

Watching Sentry for acme/storefront…The real workspace · click around

Routing nowBOT-42 · Implement

Claude · alex@14%

Claude · ops@62%

Codex91%

Cursor48%

Copilot77%

API fallbackoff

Implement is running on Claude alex@, the Claude account with the most capacity left. ops@ is next; Codex, Cursor and Copilot stand by.

Four steps. One of them is yours.

01 · CONNECT

Connect what you run.

Error monitoring, backlog, repo. Read-only until you approve.

Sentry · acme-prod · read-only · GitHub · acme/storefront

02 · AUTOPILOT

Autopilot triages.

Each new error becomes an issue with a priority and an owner.

Waiting for the next event…

03 · WORKFLOW

The workflow fixes it.

Plan → Implement → Review → Verify, routed across your accounts so it never stalls.

Not started

04 · YOU

You approve.

A pull request waits with what changed and what it cost.

Nothing to approve yet

SMART ROUTING

Your subscriptions, never idle. Your runs, never stopped.

Claude Code stops when your account hits its limit. BotInc doesn't. Each run starts on the account with the most capacity left and, when it runs dry, continues on the next one. No re-login, no lost context.

Least-used firstEvery subscription gets used before any is wasted.

Switches mid-runSame context, same thread, nothing to log into.

Runs in the cloudClose the lid. The pull request is waiting in the morning.

API only if you allow itSpill over at list price + 10%, or wait for the reset.

Routing · BOT-42 · Implement1m 39s

RUN0%

Claude · alex@Max · Opus 5 / Sonnet 514% leftRunning

Claude · ops@Pro · second account62% leftNext

CodexChatGPT Pro · GPT-6 Astra91% leftStandby

CursorPro48% leftStandby

CopilotBusiness77% leftStandby

API fallbackDeepSeek · OpenRouter · list +10%—Off

09:43:02Implement needs Claude. Opened on alex@ (Max), the Claude account with the most capacity left.

THE WORKFLOW

Every fix passes the same gate.

One workflow keeps bad code out of main, whether the issue came from Sentry at 3 a.m. or from you. Watch BOT-42 go through it.

Fix & reviewDEFAULT · 5 STEPS · EDITABLE

onissue filed by Issue intake, or by you
planmodel: auto → Claude Opus 5 · output: cause, files, tests
implementmodel: auto · tests: required · branch per issue
reviewmodel: different from author · max rounds: 3
verifyfull suite · clean sandbox · red goes back
approvalhuman: required · merge: never automatic

BOT-42Plan · reading the repo

PLANRunning

Claude Opus 5 · alex@

Read before writing.

The cause, the files, the tests to add. A plan you can read and reject before a line changes.

Reading checkout/summary.ts…GATE · plan must name cause + tests

IMPLEMENTWaiting

Claude Sonnet 5 · auto

Small diff, tests included.

The change and the tests that prove it, on a branch, in a clean sandbox.

WaitingGATE · tests required · diff ≤ 400 lines

REVIEWWaiting

GPT-6 Astra · Codex

A different model reviews.

The reviewer is never the author. It reads the diff against the issue and requests changes until it is right.

WaitingGATE · reviewer ≠ author · ≤ 3 rounds

VERIFYWaiting

BotInc Cloud · 12 checks

Checks on a clean checkout.

The full suite runs on the final commit. Red goes back, not forward.

WaitingGATE · all checks green

YOUWaiting

Alex

Nothing merges itself.

The pull request waits with the receipt. Approve from chat, the call, or your phone.

WaitingGATE · human approval

Edit it, add a step, or write your own. The default is strict.

Autopilots that come switched on.

Three routines close the loop from alert to merged fix. Turn any off, or ask your Operator for a new one.

Issue intake

Every new error becomes an issue with a priority, a release and an owner. Duplicates merge.

→Issues· 4m 12s avg

Fix & review

Plan → Implement → Review → Verify. The reviewer is never the model that wrote the code.

→

Pull request· waits for you

Regression watch

After each merge, watch the error for 24 hours. If it returns, the issue reopens.

→Reopen or close· 24h

Connect what you already run.

Read-only until you approve a change.

Sentry

PostHog

Grafana

Datadog

New Relic

Honeycomb

incident.io

Linear

GitHub

GitLab

Slack

Vercel

Your models. Your subscriptions.

Connect the plans you already pay for; runs on them cost no credit. Unfiltered open-weight models from day one.

Claude

Opus plans, Sonnet builds.

$0 CREDIT

Codex

Reviews code a different model wrote.

$0 CREDIT

Unfiltered open-weight models

No provider refusals. Your workspace, your rules.

DeepSeek 4.1 Flash UncensoredGLM 5.3 Flash Uncensored+ 6 more

DAY ONE

Anything else

Any model via OpenRouter, list price + 10%.

METERED

DESKTOP · MOBILE · YOUR HARDWARE

On your desk. In your pocket.

Browser, desktop, phone. The desktop daemon drives the coding CLIs you already licence, on your own machine, for $0.00.

16 CODING HARNESSES · LOCAL RUNTIMES FREE FOREVER

Claude CodeCodexCursor AgentCopilot CLIGemini CLIOpenCodeHermesKimiQwen CodeKiro CLIAntigravityOpenClawCodeBuddyPiQoder CLITrae CLI

The tier gates scale, never features.

Every feature is on Free. Per workspace, unlimited members. 14 days of Team to start, no card.

Free

$0forever

per workspace

1fixes running at the same time

1concurrent sessions
100autopilot runs a month
1 GBstorage included

No credit card required

Pro

Team

$150/ mo

per workspace, billed monthly

Unlimitedfixes running at the same time

Unlimitedconcurrent sessions
100,000autopilot runs a month
50 GBstorage included

Cancel anytime

FREE FOREVERLocal runtimesYour own accountsEvery feature

Connect Sentry.
Wake up to pull requests.

About two minutes. Nothing merges without you.

Coding agents that never stop.Fixes that never ship unreviewed.