Step 1
Connect the alerts you already watch
Connect the error tracker or alert channel your team already watches. Choose which alerts start a mission and which only get a note.
Incident response
Early accessTurn a production alert into a tested pull request
When an alert fires, SHIP reads the error, finds the cause in your code and writes a fix. The fix comes as a pull request that passed your CI and a test.
- For
- Engineers on call for a production web app.
- Instead of
- A stack trace you read at night to find the line that broke.
Runs today: When an alert opens an issue in Linear, assign it to SHIP. In GitHub Issues, comment @SHIP on it. The fix comes back as a pull request.
Answer 4 questions. SHIP emails you when it opens to you.
How it works
Set it up once, then review fixes
Step 2
Read what the agent found first
The agent reads the stack trace, the recent changes and the code around them, and writes the cause it found on the issue. With a checkpoint on, the mission waits for your approval before any code changes.
Step 3
Approve the fix
The fix comes as a pull request that passed your CI, a code review and a test on a preview your own pipeline deploys. Nothing reaches production until someone on your team merges it.
Capabilities
Keep a person in charge of production
Pause the mission before any code changes
A checkpoint stops the mission after the investigation, so someone approves the approach before the Builder starts.
Test the fix on a preview
The Tester runs the path that failed on a preview your own pipeline deploys and attaches the browser trace as proof.
Run on your own models and keys
Every agent runs on the credentials you connect, and your provider bills inference at your own rates.
See what every incident cost
Each mission reports what it cost and how long it took, priced on the model that ran it.
Comparison
Compare SHIP with Devin and Factory
| Factory |
|
|---|---|
| Devin |
|
| SHIP | In early access
|
FAQ
Frequently asked questions
Is alert-driven incident response available today?
Not yet: the direct connection to an error tracker or alert channel is in early access. Today, assign the issue an alert opens to SHIP in Linear, or comment @SHIP on it in GitHub Issues. The agents find the cause and send the fix as a pull request.
Does SHIP deploy the fix to production?
No. The fix waits as a pull request until someone on your team merges it, and then your own pipeline deploys it. Auto-merge stays off unless you turn it on.
When is Devin or Factory the better fit?
Factory fits a team that needs this today: its Alert Responder examines Sentry, Datadog or PagerDuty alerts as they fire, as its automations page describes. Devin fits a team that wants an agent to triage and sort alerts and backlog items and suggest approaches. SHIP fits when each fix must pass your CI, a code review and a test, on the harness and model you choose.
What does it cost to send SHIP an alert?
Each alert issue you send to SHIP is one mission. One credit covers up to 3 agent-hours, and each further 3 agent-hours or part of them uses one more. Credits come back when a mission ends in failure or needs attention with no pull request, and your own provider bills model usage.
Answer your next alert with a pull request
Join the early-access list, and assign alert issues to SHIP while you wait.
Join the early-access list
Answer 4 questions. SHIP emails you when it opens to you.