Product2 min read

Introducing Gitea support

Bring your agents into Gitea issues, pull requests, and releases.

AgentConnect now supports Gitea (GitHub and GitLab are already supported). Bring Claude Code, Codex, and your other agents into issues, pull requests, and releases on gitea.com or a self-hosted Gitea instance.

Choose which repositories an agent should follow and when it should start: on new issues or pull requests, on updates, or when mentioned. Several agents can work in the same repository. Mention an agent by name to give it a task, or request the bot as a pull request reviewer.

You can also filter issue and pull request triggers by labels, such as bug or needs-review. An item must carry at least one matching label; with Any update, adding a matching label can start the agent.

Release events can start an agent too. Choose Published for new releases or Any update for publishes and edits, then give the agent a task such as summarizing changes or assessing what a new version means for your project.

Agents reply in the issue or pull request. For code review, they can comment on specific lines, approve a pull request, or request changes. A commit status on the pull request shows the agent’s run state and links to its session.

An agent requests changes on a Gitea pull request and leaves a comment on the affected line.

With Jev-powered Decisions, you can choose which agents handle each issue or pull request based on its content, including the diff when available. For example, route bugs to a triage agent and security-related pull requests to a security reviewer.

For cross-model review, set the reviewer's Runtime to By decision and choose PR author. Jev judges who wrote the pull request, and your rules choose the runtime and model for the new review session: Claude-written PRs can go to Codex, and Codex-written PRs to Claude. Existing sessions keep their chosen runtime and model.

Runtime set to By decision with the PR author Decision: claude maps to GPT-5.5, codex and grok map to Opus, and Sonnet is the fallback.
Example rules for cross-model review: Claude-written PRs use GPT-5.5; Codex- or Grok-written PRs use Opus. Sonnet is used when no rule matches or the evaluation is unavailable.

Gitea 1.23 or later is required. For setup and configuration, see the Gitea integration guide.

Resources