Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Firmament MCP server what's our approved process for handling refunds?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Your team's shared knowledge, available to every agent.
Firmament holds what is true about your company and how work actually gets done there: the decisions, the numbers, the gotchas, and the exact steps that work in your organisation. None of that is in any model's training data, so an agent without it guesses, or works from a document that went stale months ago.
This server exposes Firmament to any MCP client over a hosted HTTP endpoint. There is nothing to install and nothing to run locally.
Two tools.
ask looks up what your team already knows, before the agent acts or
answers. It covers both kinds of knowledge: facts about your organisation (its
products, customers, metrics, status, decisions) and how work gets done in it
(conventions, prerequisites, the step that is not in the README).
submit records what the agent learned on the way out: a new fact, a
decision and its reasoning, a number, what failed and why, or the full recipe
for a multi-step process. That goes back to the shared knowledge base, so the
next session and every teammate's agent starts from it.
The result is that knowledge accumulates as work happens, instead of being written up later by whoever has time.
Related MCP server: HiveMind
The server is remote and speaks streamable HTTP. Point your client at:
https://mcp.getfirmament.com/mcpAuthentication is OAuth 2.1 with dynamic client registration, so most clients need nothing beyond the URL. You sign in once in your browser when the client first connects, and it handles the rest.
Add it from the Extensions panel by searching @mcp, or add the URL directly to
your MCP configuration as an HTTP server.
Any MCP client that supports remote HTTP servers works. Firmament also ships first-class packages for several agents, so you may prefer those:
Agent | Install |
Claude Code |
|
Gemini CLI |
|
Cursor, Codex, Copilot CLI | see firmament-plugin |
You do not need to prompt the tools by name. The packages above ship a standing instruction so the agent consults Firmament on its own. With a bare MCP connection, the tool descriptions carry the same guidance.
In practice it looks like this:
Before a deploy, the agent asks how releases are actually done here and gets your runbook, including the manual step missing from the README.
Asked about pricing, a discount or a refund, it checks your team's approved answer before replying to a customer.
After a long debugging session, it records the root cause and the fix, so nobody loses a day to it again.
After a multi-step migration, it submits the recipe with the exact commands, so the next person runs it in one pass.
Firmament is a hosted product. Sign up at getfirmament.com. Knowledge is scoped to your workspace, and the server resolves every call to a workspace and user from the OAuth token, so one team's knowledge is never visible to another.
Product: https://getfirmament.com
MCP endpoint: https://mcp.getfirmament.com/mcp
Registry entry: com.getfirmament/firmament
This server cannot be deployed
Shared knowledge base for AI agents. Semantic search across agents, no setup required — just a URL.
Your company's brain for AI agents. Cited, permission-aware knowledge across every system.
Shared, permission-aware company context for AI agents, with provenance, approvals and audit.
Shared company knowledge, workflows, and connected apps for the AIs your team already uses.
