Rulith documentation
Run AI agents on your computer, connect them to governed tools, and inspect the evidence behind their work.
Rulith has three places to work: the local Rulith workbench for conversations and Workers; Console for your account, Agents, capabilities and permissions; and the Board, where the service evaluates facts, rules, Actions and Case outcomes.
Choose a starting point
| You want to… | Start here |
|---|---|
| Run your first Agent | Install and sign in |
| Set a default model or switch Agents | Agents and Workers |
| Work with files on your computer | Files and permissions |
| Turn a document into a checked capability draft | Document to draft |
| Install or configure a capability | Capability configuration |
| Understand evidence and completion | Core concepts |
| Diagnose a blocked step | Troubleshooting |
What is available
These guides cover the published Rulith 0.8.17 workbench and the hosted service checked on September 28, 2026. Local sign-in, multiple Agent profiles, local files and the document assistant are available. The protocol reference follows the deployed Java service.
Candidate protocol revisions are not instructions for the hosted service. Upgrade guidance changes when the corresponding server and client have shipped together.
Rulith is an English-language public beta. Platform payments are inactive. Your chosen model provider may charge separately, and remote models receive the material you authorize for that destination.
Keep the responsibilities clear
The Agent chooses what to investigate or request. Rulith evaluates the request against the current program and evidence. A Worker runs permitted tools near your data and reports what actually happened. A Case records one piece of governed work; it is separate from a chat conversation.