NewCapy V2 is here

Work that
starts itself

Set up recurring or event-driven agent workflows that run automatically. Sentry alert comes in, PR goes out. Linear issue assigned, work begins. No human needed until review.

Capy resting
Threads
JSJustin SunAcme, Inc.
AutomationsSentry regression → root cause
View runsSave
Sentry regression → root cause
Active
Frontend
MCMia Chen
MCBy Mia Chen
Triggers
Runs when any of these conditions are met
Issue changedincapy-webandwith actionCreated, Reopened or regressedandwith levelError, Fatal
Instructions
What Capy does when the automation triggers, trigger context is added automatically
Investigate the Sentry issue using the event payload, stack trace, release/environment data, suspect commits, and attached repositories. Identify impact, likely root cause, affected code path, and whether this is new, regressed, or already known. Implement and verify a focused fix when confidence is high; otherwise produce a concise investigation note with the missing evidence and recommended owner.
Fable 5Extra high · Fast
Only run when
A fast model checks each event against this and skips runs that don't match
Run when the issue has enough stack trace or event context to investigate a product defect. Skip noisy low-severity events without actionable code context.
Limits
Max runs per day
Extra events are skipped and stay visible in run history
No limit
Machine size
Compute for each run's machine, project default when unset
HyperProject
Usage
Runs and spend per day
$17.2132 runs in the last 30 days
Spend
Runs
Volume
Durable files available to every run
Sentry regression → root causeAutomation
3 files · 1.2 MB
Updated 2h ago

Start agents for any workflow

Trigger agents from events in Sentry, Slack, Linear, or GitHub, or on a schedule. Each trigger starts a full agent run and ends as a task you can review.

Event-based triggers

Sentry issues, GitHub pull requests, Linear tickets, Slack messages, or any webhook payload.

Scheduled jobs

Daily dependency updates, a weekday standup digest, a Friday changelog. Presets or raw cron, in your timezone.

7:00 AM
Dependency updates
8:00 AM
9:00 AM
Standup digest
10:00 AM
Flaky test triage
11:00 AM
12:00 PM
Security scan
1:00 PM
Changelog draft

API & webhooks

Launch agents from your own code. Create and run automations over the API, with whatever context your systems already have.

GET

/v1/automations

POST

/v1/automations

GET

/v1/automations{automationId}

POST

/v1/automations{automationId}/run

GET

/v1/automations{automationId}/runs

POST

/v1/automations{automationId}/pause

POST

/v1/webhooks{webhookId}

GET

/v1/threads{threadId}

POST

/v1/threads{threadId}/message

A full environment on every run

Each run gets its own machine with your environment variables, a durable volume, and the limits you set.

Environment variablesAdd
DATABASE_URL••••••••••2d
SENTRY_DSN••••••••••5d
LINEAR_API_KEY••••••••••1w
NPM_TOKEN••••••••••3w
Volume
Sentry triageAutomation7 files · 4.2 MB
Durable across runs
triage-playbook.md6 KB
known-issues.json41 KB
repro/3 files
scripts/bisect.sh2 KB
Updated 3h ago by a run
Limits
Max runs per dayExtra events are skipped
24
Machine sizePer-run compute
MediumProject

Automations teams run

A nightly batch, a Monday schedule, and a Slack thread. Each one ends as a task you review.

Tasks6
Task 1Bump next 14.2.3 → 15.0.1Fable 5capy#146
Task 2Bump react and react-dom to 19.1.0Fable 5capy#147
Task 3Bump prisma 5.22.0 → 6.2.1Fable 5capy#148
Task 4Bump zod 3.23.8 → 3.24.1Fable 5capy#149
Task 5Bump bullmq 5.12.0 → 5.34.2Fable 5Running
Task 6Bump tailwindcss 4.0.0 → 4.1.3Fable 5Running

Dependency updates

Every night, Capy bumps what is outdated across your repos, runs the tests, and opens one PR per package with a note on what moved.

On a scheduleAmerica/Los_Angeles
Mondays at 9:00 AM
2026-09-01.mdxchangelog+64
1+## Week of Aug 25
2+
3+### Highlights
4+- Rename a thread with ⌘⇧R (#4819)
5+- Slack replies no longer post empty blocks (#4811)
6+- Folders in the sidebar (#4790)
7+
8+### Dependencies
9+- Two nightly batches landed (#4802, #4805)
Pull request openedDraft
docs: changelog for the week of Aug 25 #143

Weekly changelog

Every Monday, the PRs merged since the last changelog become a docs PR with the release notes drafted.

#bugs
Mira
Mira2:41 PM
the shortcut hint overlaps the composer on narrow widths, anyone else?
Capy
1 replyLast reply today at 2:43 PM
Jon
Jon3:12 PM
export does nothing on Safari 18
Capy
3 repliesLast reply today at 3:15 PM
Ana
Ana3:40 PM
build failing on main since #4802
Capy
1 replyJust now

Slack bug triage

A bug lands in a channel. Capy reproduces it, ships the fix, and replies in the thread with the PR.

FAQ

Triggers, pull requests, failed runs, and controls.

  • What can trigger an automation?

    Schedules, GitHub events, Slack messages, Linear updates, and API or webhook calls can all start an automation. Each trigger creates a fresh Capy task with the prompt and environment you configured.

  • How is an automation different from a regular task?

    A regular task starts when you ask for it. An automation starts when its trigger fires, then runs as the same kind of reviewable Capy task with its own thread, status, files, and result.

  • Can an automation open a pull request?

    Yes. Every run gets a full development environment, so Capy can edit code, run tests, open a pull request, and keep working through CI failures and review comments.

  • What happens if a run fails?

    The failed run stays in the automation history with its full thread and output. You can inspect what happened, update the prompt or environment, and run it again.

  • Can I pause an automation or run it manually?

    Yes. Pause an automation without losing its configuration or history, turn it back on whenever you want, or start a run immediately instead of waiting for the next trigger.

Set it and ship it

A trigger and a prompt. Your first automation takes under a minute to set up.

Get startedContact sales
Capys jumping through a portal