Documentation
Connect LithicBlue to Microsoft Copilot
Before you start
What you need
A LithicBlue account on a paid plan, and your MCP endpoint, which is your regional URL with /api/mcp/mcp appended (e.g. https://eu.lithicblue.com/api/mcp/mcp). See the MCP connector page for your exact URL. On the Microsoft side you'll use Copilot Studio, and the agent needs generative orchestration turned on to use MCP.
Add the MCP tool to an agent
- In Copilot Studio, open the agent you want to extend (or create one) and make sure generative orchestration is on.
- Go to the agent's Tools page and choose Add a tool → Model Context Protocol → New tool → MCP.
- Enter your MCP server URL (your endpoint above).
- For authentication, use OAuth 2.0; LithicBlue supports dynamic client registration, so the dynamic-discovery option configures it for you. Authorise and sign in to LithicBlue, then review and approve the access request on the Approve access screen, which names the connector, the host it connects back to, and what it will be able to do. Check it names Copilot before you continue.
- Select Add and configure. LithicBlue's tools are now available to that agent.
Tools stay in sync
Copilot Studio reflects changes to the connected server automatically, so new LithicBlue tools appear (and removed ones disappear) without re-adding the connector.
How to track your projects in Copilot
Ask the agent to list your LithicBlue projects, then summarise scope fulfilment and the open data gaps on any project. It reads your scope analysis and data-gaps directly, so you can check where a piece of work stands from inside Copilot.
How to verify AI outputs with Copilot
Ask the agent to read a deliverable's claims and tell you which are verified, contradicted or still unverified against your sources, and to author new claim highlights anchored to the exact passage. This is how you fact-check AI-generated and human-written deliverables with traceable evidence. The full tool list and example prompts are on the MCP connector page.