MCP server

Bidderops exposes a Model Context Protocol (MCP) server so AI assistants — such as Claude or Cursor — can securely read and act on your bid pipeline on your behalf. Instead of copy-pasting between tools, you can ask your assistant questions like “what deadlines are coming up this week?” and have it answer from live Bidderops data.

How a connection works

The MCP endpoint is an authenticated connection bound to your account. At a high level:

Your assistant connects to the Bidderops /mcp endpoint.
Because it is unauthenticated, the server points the assistant at its authorization server (standard OAuth 2.1 discovery).
The assistant registers itself and opens an authorization page where you log in.
After you approve, the assistant receives a short-lived access token tied to your session.
Every tool call the assistant makes runs as you, under the same data isolation that applies in the app — so it can only ever see your organization’s data, and only what your role permits.

You can review and revoke connected assistants at any time in Settings → Integrations.

Available tools

The server exposes a focused set of read and write tools:

Tool	What it does
`list_opportunities`	List opportunities, optionally filtered by status or text.
`get_opportunity`	Fetch one opportunity with its notes, documents, milestones, team members, and outcome.
`upcoming_deadlines`	List active opportunities with a submission deadline within the next N days (default 14).
`pipeline_summary`	Headline metrics: count by status, open count, open pipeline value, and win rate.
`summarize_losses`	Group lost bids by loss reason and collect lessons-learned notes.
`add_note`	Add a timestamped note to an opportunity (requires an editor role).

Reads are available to any role; writes such as add_note require an editor role, and that boundary is enforced by the database — not just the tool — so a viewer’s assistant cannot write even if asked to.

Self-hosting the connector

A standalone MCP connector is also available to run yourself, either as a local subprocess (stdio) or as a small hosted service with the same OAuth flow. See the project’s docs/mcp-remote.md and mcp/README.md for deployment details.