Forgejo

• “Which repositories can I see?”
• “Which issues are open in the Fachverfahren repository?”
• “Summarize the comments on pull request 12.”
• “Search all repositories for open issues with the keyword Import.”

Forgejo, Gitea and Codeberg speak the same interface, so there is exactly one connection for them. Codeberg is a hosted Forgejo instance; you simply enter codeberg.org as the address.

1. 01In your Forgejo, Gitea or Codeberg instance open user settings.
2. 02Choose Applications and there Generate New Token.
3. 03Give a name. From Gitea 1.20 or current Forgejo versions choose read-only rights: read:repository, read:issue, read:user.
4. 04Generate the token and copy it immediately. It is shown only once.

1. 01In anymize open Apps, then Connectors, and choose Forgejo.
2. 02Enter your instance address and paste the token from the section above. Older instances ignore the read-rights choice; there the token carries your full rights.
3. 03Set which areas the assistant may use: account, repositories, issues, pull requests.

No. There is no tool for create, comment, close or merge. The connector only sends read requests, regardless of how much your token would allow.

That is exactly why the guide above says you should choose pure read rights from Gitea 1.20 on: on older instances read-only is our commitment, on newer ones it is also that of your own server. Two locks are better than one.

Forgejo and Gitea are open source and run on a server you operate yourself. You give the address, anymize talks only to that address, and the token stays encrypted in your workspace.

• Do not register an application with the vendor, no approval process.
• Codeberg is the hosted variant for anyone who does not want to run their own server.
• Disconnecting removes access immediately.

• Source code is not read, only repository details, issues, pull requests and comments.
• A list returns at most 50 entries per page; many instances also cut to their own maximum.
• Releases, pipelines and wiki pages are not included at this stage.

What goes through is anonymized: usernames, real names and email addresses from issues and comments are replaced by placeholders before an AI model summarizes the history.