Monk Fix
Enable the AI agent that opens a pull request with a fix when your CI job fails.
When a CI job fails, Monk Fix reads the logs, identifies the root cause, edits the files it believes are at fault, and opens a pull request with the fix.
Where to find it
Monk Fix appears in your dashboard under Agents as the Debug Agent. It is opt-in per repository - nothing runs until you enable it.
Enable Monk Fix
Open Agents
In your Monk CI dashboard sidebar, go to Add Ons → Agents.
Connect the MonkCI Agent App
Click Connect / Install and authorize the MonkCI Agent App on your GitHub organization. Grant it access to the repositories you want agents to work on.
Return to the dashboard. Click the refresh icon next to the connection status if it has not updated.
This is a separate GitHub App from the Monk CI runners app. Repository access is controlled on GitHub - use Manage on GitHub to change which repositories the app can reach.
Open the Debug Agent
In the Agent catalog, select the Debug Agent card. Its trigger reads Runs when a CI workflow fails.
Enable the repositories you want it to act on
Toggle each repository on, or use the search box and select every repository shown at once.
A repository that is off is never touched. Changes save as you make them.
What Happens When a Job Fails
No configuration is added to your workflow files. Once a repository is enabled:
- A CI job in that repository fails.
- Monk Fix analyzes the failed job's logs and identifies the root cause.
- If it can produce a fix, it pushes a branch and opens a pull request.
- The result appears on the failed job in Run History.
Review before merging
Monk Fix opens pull requests for you to review. It never pushes to your default branch and never merges its own PR.
Read the Fix
Open the failed run from Run History and find the failed job. Monk Fix reports inline on that job:
| State | What it shows |
|---|---|
| Debug Agent is analyzing this failure… | The run is in progress. |
| Debug Agent opened a fix | The root cause, the error class, the files edited, the branch, and a View PR button. |
| Debug Agent could not generate a fix | The reason it stopped. |
Each completed fix is tagged with an error class:
| Class | Meaning |
|---|---|
| Application bug | The failure is in your application code. |
| Dependency issue | A dependency caused the failure. |
| CI / configuration | The workflow or its configuration is at fault. |
| Environmental | Something outside your code failed - review before assuming a code change is needed. |
Click View PR to review the change on GitHub, then merge it like any other pull request.
Track Agent Runs
On the Agents page:
- Recent activity lists the latest runs across all agents.
- All runs is the full history. Filter by agent to see only Monk Fix.
| Status | Meaning |
|---|---|
| Pending / Running | The agent is working. |
| Completed | A fix pull request was opened. |
| No fix | The agent ran but could not produce a fix. |
| Failed | The run errored. |
Select a run to open the pull request it produced.
Credits
Agent runs consume credits. The credits left chip at the top of the Agents page shows your remaining balance - your free grant plus any purchased credits. The chip warns you when the balance runs low.
Other Agents
The Agent catalog also lists other agents. A card marked Not yet available is coming soon; one marked with a plan name requires an upgrade before you can enable it.
Troubleshooting
| Problem | What to do |
|---|---|
| "The MonkCI Agent App isn't connected" | Click Connect / Install and authorize the app on your organization. |
| App shows as suspended | Re-enable the app on your GitHub organization, then refresh the status. |
| A repository is missing from the list | Use Manage on GitHub to grant the MonkCI Agent App access to it. |
| No fix was opened for a failed job | Check the run status. No fix means the agent could not produce one - read the AI Log Summary for the diagnosis instead. |