Teams · Engineering
Agent Serve agents triage new errors, gather incident context, sort the backlog and draft release notes. Engineers review, merge and decide.
Agent Serve, the AI workspace where teams build, deploy, and manage AI agents, gives engineering teams agents for error triage, incident context, issue sorting, pull request summaries and release notes, connected to GitHub, GitLab, Linear, Jira, Sentry, Datadog and PagerDuty.
At a glance
The problem it removes
New errors land in a channel nobody owns. During an incident, someone gathers logs, deploys and the timeline by hand. Release notes are written from memory on release day.
Agents to build
Each agent is built in Agent Serve and works with the code host, issue tracker and observability tools the team already uses.
When Sentry raises a new issue, an Agent Serve workflow reads the stack trace and recent deploys, and files a Linear or Jira issue with an owner.
When PagerDuty or incident.io opens an incident, an Agent Serve Relay gathers recent Datadog monitors, logs and deploys and posts a summary to the incident channel.
An Agent Serve Relay reads open issues, groups duplicates and proposes labels and priorities for the team to accept.
An Agent Serve workflow reads the pull requests merged since the last release on GitHub or GitLab and drafts release notes for review.
An Agent Serve agent creates Plane work items from triaged errors and adds them to the current cycle.
Examples show what teams can build in Agent Serve. They describe patterns, not named customer deployments.
How it runs
Agent Serve gathers the signals while the responder gets to a laptop. People run the incident.
A PagerDuty Incident Triggered event starts the Relay’s task.
The Relay reads Datadog monitors and logs for the affected service.
It lists the deploys and merged pull requests from the last hours.
A summary with links goes to the incident channel in Slack.
Integrations
Agent Serve connects to code hosts, issue trackers, error tracking, monitoring, on-call and deployment tools.
Browse integrationsIn each tool
Agent Serve reads pull requests, files and commits, comments on reviews, and starts when an issue or pull request opens.
PR opened → summary comment
Agent Serve creates, updates, labels and comments on issues, starting when an issue is created or changes.
Issue created → labelled
Agent Serve reads issues, events and incidents, adds notes, and starts when an issue or incident is created.
Incident triggered → context posted
Agent Serve queries metrics and logs, lists monitors and incidents, and can create a downtime when a person approves.
Deploy window → downtime
Agent Serve creates and updates Plane work items, cycles, modules and comments.
Triaged error → work item
In the workspace
Workflows start on Sentry, GitHub or PagerDuty events and file the work.
A Relay collects signals during an incident and keeps the timeline.
Knowledge bases sync runbooks and architecture docs from Confluence or GitHub.
Logs trace every agent run block by block.
In depth
The Agent Serve workflow builder shows each block, its inputs and outputs, and lets you test a run before you deploy.
Agent Serve Chat answers from synced runbooks and docs, and links the page it used.
Agent Serve logs every run, and the Relay audit view lists permitted and refused actions with the rule behind each.
Before and after
| Work | Before | With Agent Serve |
|---|---|---|
| New errors | BeforeA channel of alerts nobody owns. | With Agent ServeAn issue filed with the trace and an owner. |
| Incidents | BeforeThe first ten minutes spent gathering links. | With Agent ServeA briefing in the channel when the responder arrives. |
| Release notes | BeforeWritten from memory on release day. | With Agent ServeDrafted from merged pull requests and edited. |
Governance
An Agent Serve Relay uses only the integrations and resources on its allowlist, and writes wait for approval unless you delegate them.
Relays never send email or book meetings without an approval, under any autonomy setting. An approval covers the exact payload the reviewer saw and expires if nobody decides.
Workflows run from a deployed version, so a change is reviewed before it reaches production runs.
Every workflow run records each block’s inputs, outputs and errors, so a reviewer can trace how a result was reached.
Run Agent Serve in your own cloud or data centre, next to the systems it reads.
Questions
The GitHub integration can merge a pull request if you build a workflow to do it. Most teams keep merging with engineers and use agents for summaries and triage.
Yes. Agent Serve agents can be built visually in the workflow builder, conversationally in Chat, or programmatically with the API.
Agent Serve has integrations for GitHub, GitLab, Bitbucket, Azure DevOps, Linear, Jira, Sentry, Datadog, PagerDuty and incident.io, among others in the integrations catalog.
Last updated
Enterprise access
Agent Serve is available to teams by invitation. Tell us what you want to build and we will set up your workspace with you.