Skip to content

C:\> Documentation

Integrations

Contents

Connect your team's tools and the agent can pull them into an investigation: searching existing issues, filing new ones, and reading your wiki. Connections live under Integrations in the app: one card per provider, each asking for the credential that provider issues.

Issue trackers and docs sources
Connect your issue tracker and wiki: the agent searches both while it investigates, and files a follow-up issue only when you approve one.
Issue trackers and docs sources
Screenshot of Issue trackers and docs sources. Connect your issue tracker and wiki: the agent searches both while it investigates, and files a follow-up issue only when you approve one.
Connect your issue tracker and wiki: the agent searches both while it investigates, and files a follow-up issue only when you approve one.

Fill in what the provider needs, hit Test Connection to check it against the provider before you commit to it, then Save. The badge on each card tells you whether that provider is currently connected, and a provider only appears to the agent as a tool once it is.

Issue trackers

Connect an issue tracker and the agent can search existing issues and file new ones as part of an investigation. Codepanion ships four providers. Jira needs your site URL, the Atlassian account email the token belongs to, the project key new issues are filed under, and an API token; a save or test with any of those missing is refused and names the field. Linear wants a personal API key, GitHub a personal access token, and Salesforce an OAuth access token. Salesforce is read-only: the agent can search it, but Codepanion never writes there.

Once connected, the agent gains these tools:

  • search_issuesFind related issues so an investigation can build on what's already known.
  • create_issueFile a new issue, matching your tracker's house style, when an investigation surfaces something worth tracking. Offered only for providers that can create issues, so it never appears for Salesforce.
  • propose_changeTurn a proposed fix into a unified-diff patch. On GitHub it is attached to the issue as a comment, and a one-line change also opens a draft pull request; on the other trackers the patch is handed back in the thread.

Both of those write to your tracker, so neither ever runs on its own: the agent shows you exactly what it wants to send and waits for you to approve it. See approvals & write actions.

Checking the tracker before writing to it
The agent searches your issue tracker before it files anything, so a known problem gets linked to the issue that already exists.
Checking the tracker before writing to it
Screenshot of Checking the tracker before writing to it. The agent searches your issue tracker before it files anything, so a known problem gets linked to the issue that already exists.
The agent searches your issue tracker before it files anything, so a known problem gets linked to the issue that already exists.

Searching first is much of the point of connecting a tracker. An investigation that finds the problem already on file links to it instead of duplicating it, and one that finds nothing drafts the new issue out of what the investigation established rather than out of the sentence you typed.

Docs sources

You can also connect a docs source: link Confluence (your site URL, Atlassian account email, and an API token) and the agent can search your team's wiki (runbooks, architecture notes, postmortems) and read the pages it finds:

  • search_docsSearch your documentation for relevant pages; each hit comes back with its title, link, and an excerpt.
  • read_doc_pageRead a page's full content when an excerpt looks relevant, and cite it in the findings.

Note

Integration credentials are stored per tenant and used only for that tenant's investigations. More providers are on the roadmap; tell us which tracker you use.

Need help getting set up?

We do the first setup on a call with you: the CI step, the connection string, and a first investigation against your real code.