Issue Tracking

Connect Linear and Sentry, then turn an issue into an agent run from Agent Home

Overview

Connect your issue tracker to Agentastic and start an agent straight from a bug report. Pick an issue in Agent Home and Agentastic hands the agent the issue's title and a direct link as context — so it begins with the real problem in hand instead of a prompt you have to write from scratch.

Two trackers are built in:

  • Linear — pull in an issue from your team's backlog.
  • Sentry — turn a production error into a fix, with the agent pointed straight at the offending issue.

Both store their credentials securely in the macOS Keychain and surface the linked issue across the sidebar, Kanban board, and chat.

Connecting an Account

Issue trackers live alongside your Git accounts in Settings → Accounts. Each tracker has an Enabled toggle and a row you can click to open its details.

Linear

  1. Open Settings → Accounts and click the Linear row.
  2. Paste a personal API key into Paste API key. You can create one at linear.app/settings/api.
  3. Click Save. Agentastic validates the key and shows Connected.

To remove the key later, open the row and click Disconnect….

Sentry

  1. Open Settings → Accounts and click the Sentry row.
  2. Paste a personal auth token into Paste auth token. You can create one at sentry.io/settings/account/api/auth-tokens/.
  3. Click Save. Agentastic validates the token and loads your organizations.
  4. If you belong to more than one Sentry organization, pick which one to search from the Organization dropdown. Your choice is remembered.

To remove the token later, open the row and click Disconnect….

Linking an Issue in Agent Home

  1. Type your task in the Agent Home prompt as usual.
  2. Click the + button below the prompt and choose Linear Issue… or Sentry Issue… (these appear once the tracker is connected).
  3. Search for the issue — Sentry defaults to your most recent unresolved issues — and click one to link it.
  4. The linked issue appears as a chip above the prompt showing its identifier and title. Click the ✕ on the chip to unlink it.
  5. Send the prompt to the agent (any mode — normal, auto-approve, plan, explore, or chat).

What the Agent Receives

The linked issue is prepended to your prompt so the agent has the context before it starts. For example:

Sentry Issue: PROJECT-1A2 - NullReferenceError in checkout flow
URL: https://your-org.sentry.io/issues/12345/

<your prompt>

Linear issues are injected the same way, with the issue identifier, title, and URL.

Where Linked Issues Appear

A linked issue stays attached to the worktree it launched, so you can pick the work back up later:

  • Navigator sidebar — the issue identifier shows next to the worktree.
  • Kanban board — the identifier appears as a badge on the worktree's card.
  • Chat — chat sessions can carry a linked issue and display it as a chip.
  • Scheduled Tasks — schedule an agent to run against a specific issue.

Tips

  • Sentry → fix loop — link the failing Sentry issue, run an agent in plan mode first to scope the fix, then auto-approve once the plan looks right.
  • Keep the link — because the issue follows the worktree, you can close Agentastic and return to exactly the right context.
  • Multiple trackers — Linear and Sentry can both be connected at once; link whichever fits the task.