Home

tralha is self-hosted storage for the files AI agents produce: screenshots for pull requests, reports, logs and notes. Agents reach it through an MCP server or a REST API, people through a web panel, and anything can be shared with a revocable public link.

The README covers installation and the quickest path to a working setup. These pages explain how each part works and why it was built that way.

Pages

• Architecture: the two ports, the request path and where data lives on disk.
• Buckets, keys and links: the data model, permission scopes, public links and uploads.
• MCP server: the tools an agent gets and how binary files reach the server.
• Admin panel: what the web panel does and how it is protected.
• Cleanup and compression: cleanup rules, quotas and WebP compression.
• Security model: the threats tralha expects and the defence against each.
• Deployment: running on a VPS, behind a VPN, with Postgres, plus backups and upgrades.
• Reference: environment variables, CLI commands and REST endpoints.

Version

These pages describe tralha 0.2.0. Changes between versions are listed in CHANGELOG.md.