Skip to main content

SchemaBounce for Linear

The SchemaBounce Linear app puts your hosted SchemaBounce agents inside Linear. Assign an issue to the app, or mention it in a comment, and the agent you chose for that team does the work in your SchemaBounce workspace and posts its answer back on the issue.

The app is in developer preview. It is built and running, and we are still proving it with early workspaces before we call it generally available.

Publisher: SchemaBounce · schemabounce.com · Contact us

What it does​

  1. Someone in Linear assigns an issue to the SchemaBounce app, or mentions it in a comment.
  2. Linear starts an agent session and sends it to SchemaBounce.
  3. SchemaBounce posts a short note on the issue saying it has started.
  4. The agent mapped to that Linear team runs in your workspace, with the same tools, approval rules and spending limits it has everywhere else.
  5. The agent's answer is posted back into the same agent session on the issue. If the run fails, the session gets an error message instead.

A follow-up comment in the same session starts a new run with the new message.

Install it​

You need a SchemaBounce workspace and permission to manage its agents. Linear will ask someone with permission to install apps in your Linear workspace to approve the install.

  1. In SchemaBounce, open the Orchestrators page and find the Linear agent card.
  2. Select Connect Linear. Linear opens and shows the permissions below. Approve them.
  3. Back in SchemaBounce, pick a default agent. Issues from any Linear team go to it unless you map that team to a different agent.
  4. Optional: map each Linear team to its own agent.

If no default is set and a team is not mapped, the app replies on the issue that no agent is configured for that team and does nothing else.

Permissions it asks for​

Linear scopeWhy the app needs it
readRead your organization name and team list so you can map teams to agents.
writePost the agent's status notes and answers into the agent session.
app:assignableLet people assign issues to the app.
app:mentionableLet people mention the app in comments.

The app installs as its own Linear user, so its activity is labeled as the app and never as one of your people.

What it reads, and what it keeps​

At install: your Linear organization's ID and name, the app's own user ID, and your teams' IDs, names and keys.

On each agent session: the context Linear sends with the session (the issue and the comment that started it) or the text of a follow-up message, plus the session ID, the issue ID, its identifier (such as ENG-123) and its URL.

What SchemaBounce stores:

  • Your Linear access and refresh tokens, encrypted. Only core-api decrypts them, to call Linear for your workspace.
  • Your team-to-agent mappings and your default agent.
  • For each session, the session ID, issue ID, issue identifier, issue URL and the ID of the agent run it started, so the answer goes back to the right issue.
  • The delivery ID of each message from Linear, so a message Linear sends twice runs once.

The issue text is encrypted before it enters the queue that hands work to the agent, and it stays in your workspace's run history like any other agent run. We never sell it, never train AI models on it, and never use it for ads. Your data is yours has the full list of promises.

What it writes back to Linear​

Only messages in the agent session it was given: a "started" note, the agent's answer, or an error. The app does not edit issues, change their status or assignee, or create new issues.

How messages are checked​

Every message from Linear is signed. SchemaBounce checks the signature against the exact request body, rejects any message more than 60 seconds old, and confirms it was sent for the SchemaBounce app before it reads anything else. A message that fails any check is dropped.

Cost​

A run started from Linear is an ordinary agent run in your workspace. It counts toward your plan's usage and any spending limit you set on the agent, the same as a run started from chat or a schedule. See pricing.

Remove it​

From SchemaBounce: open the Orchestrators page, then Disconnect on the Linear agent card. SchemaBounce revokes the token at Linear and deletes its stored copy. Agent sessions on Linear issues stop reaching your agents at once.

From Linear: remove the app from your Linear workspace settings. Its tokens stop working, so no new work reaches your agents. Disconnect it in SchemaBounce too, to clear the connection there.

Common questions​

Which agent answers a Linear issue?​

The agent mapped to the issue's Linear team. If the team has no mapping, the workspace's default agent answers. If there is neither, the app posts a message saying no agent is configured and does not run anything.

Can the agent change my Linear issues?​

No. The app can only post messages into the agent session it was given. It cannot edit, move, reassign or create issues.

Does SchemaBounce train AI models on my Linear issues?​

No. SchemaBounce never trains models on your data, never sells it, and never uses it for ads.

Do approval rules still apply to runs started from Linear?​

Yes. A run started from Linear goes through the same approval rules, tool access and spending limits as any other run of that agent.

How do I stop the app?​

Select Disconnect on the Linear agent card in SchemaBounce. That revokes the token at Linear and stops new sessions from reaching your agents.