MCP overview
Connect your AI tools to shared Orka knowledge through the Model Context Protocol.
How the connection works
Orka’s remote MCP server exposes tools for discovering authorized projects, reading accepted context, finding shared sessions, and preparing handoffs. Your connection uses a browser authorization flow and the read-only orka.read scope.
Connect your host
- 1
Choose the right integration
Use the remote package for reading shared context in ChatGPT. Use a compatible desktop package in Codex or Claude Code when you need local selected sharing.
- 2
Authorize the Orka service
Complete the browser consent flow using an existing Orka account. The service must have MCP enabled and support the package you are using.
- 3
Select a project
List accessible projects and choose the correct project and repository source. A connected account does not grant access to every team’s data.
- 4
Read and verify
Retrieve accepted context or a shared session and check its sources, dates, and revisions.
Try your first request
Use Orka to list my projects.
For the project I select, find accepted context relevant to onboarding.
Cite the source and explain when it was last updated.Enable selected local sharing
Local sharing requires the updated Orka app, the desktop plugin, an active connected folder, and a service advertising selected-sharing support. The local and remote accounts and service origins must match.
Explicitly choose the resource, inspect its frozen preview, and report the resulting publication state. One snapshot never grants ongoing access to future edits or turns. ENV sets are excluded.
Have a question or an idea? We’d love to hear it