Your Agents.
One working environment.
Conversations in the middle, Agents on the left, execution and evidence alongside. Configure a model once, then choose overrides where a task needs them.
A clear place for every part of the work
Local workbench
Sign in, choose an Agent, set its model and start talking. Keep its conversations, files and Worker settings together. Switching Agents keeps their identities separate.
Console
Create and enable Agents, install capabilities, configure Source bindings and inspect Cases. Account administration stays separate from the conversation.
Worker
Run permitted tools near your data. Each Agent retains its own Connection, credentials and working files, even when several run on one computer.
Board and Cases
The shared evidence graph evaluates rules and Action conditions. Cases identify pieces of work and retain their recorded outcomes.
Use your model. Keep control of material.
Choose a remote model service or a compatible local endpoint. Attach a file to the local Worker material area and authorize how it can be read. A remote model receives text you permit for that destination; local execution does not mean all content stays on the computer.
Extend behavior with capabilities
Vocabulary
What the workflow talks about.
Rules
What follows from the evidence.
Actions
What work may be requested.
Sources
Where evidence and execution originate.
A Constitution adds separate governance. Connections and credentials are deployment settings, outside portable capability content.
Start with one Agent
Install Rulith, sign in and configure its model. Add a capability when the task needs governed work.
Install and sign inBring an MCP client