Skip to main content
Agent handoff turns an alert into an issue your team’s coding agent (Claude Code, Cursor, Codex, or a GitHub Copilot agent) can pick up: the crash, where it happened, what came before it, and how to dig further with the Apsio MCP server. It is off until an owner turns it on, and Apsio never pushes code, never merges and never assigns anyone: the agent opens a pull request, and your team reviews it.

​
When it opens an issue

A handoff destination is an alert destination: rules send to it like to Slack. It takes two presets:
  • New issue (new_issue), with min_users for “a new crash with more than N users”, and optionally a release.
  • Crash-free sessions (crash_free_below): one issue when the rule fires, naming the crash groups behind the drop; when it resolves, Apsio comments on that issue and leaves it open.
Each crash group gets one issue per destination, whatever the number of rules, retries or deliveries: Apsio records the issue before it asks GitHub or Linear, and after an answer it did not get (a timeout), it looks for the issue it may have opened before trying again. At most 5 issues per project per day (1 to 50, set on the destination).

​
What the issue says

  • The numbers: users affected, occurrences, sessions, first and last seen, releases and devices.
  • The in-app frames of the latest occurrence (up to 20), with file paths relative to your project: the build machine’s part (/Users/<name>/..., a CI checkout) is removed.
  • The breadcrumbs before it in its session (screen and lifecycle names; other breadcrumbs only as their kind) and the requests in the 10 minutes before it, as templated routes with their status.
  • Links to the issue in the console and to the MCP tools (get_issue, search_sessions).
It never includes an exception message, a log body, a URL with its query, or a user, installation or session id. Everything that came from the app is shown inside code blocks, under a note that tells the agent it is data to investigate, not instructions.

​
Set it up

Only owners connect, choose and turn on; admins can turn it off.
  1. In Settings > Alerts, choose Connect GitHub or Connect Linear.
    • GitHub: install the Apsio App on the repositories you choose. It asks for Issues: write and Metadata: read only, and cannot read or write code. You can only choose a repository you can push to yourself; Apsio checks it again when the destination is turned on.
    • Linear: authorize the Apsio app. It acts as itself (issues are created by the app, not by a person) with the read, issues:create and comments:create scopes.
  2. Back in Apsio, choose the repository or the team, and a name. The destination is created off.
  3. Choose Turn on, and type where issues go (for example github.com/acme/app) to confirm. Your organization’s audit log records who turned it on, when, and where.
  4. Add a rule with the New issue or Crash-free sessions preset that sends to it.
Send test checks that Apsio can still reach the repository or the team, without opening an issue, even while the destination is off.

​
Turning it off

  • Turn off on the destination stops it at once: nothing pending is sent.
  • Disabling or deleting the rules that send to it, or deleting the destination, stops it too.
  • Uninstalling the App on GitHub (or removing the repository from it), or revoking the app in Linear, revokes the destination: Apsio sends nothing more until an owner connects again.
The issues Apsio opened stay in your tracker; Settings > Alerts lists them.