Developers
TenderLedger MCP
Public notices give you the facts. TenderLedger gives AI agents the next move — without raw database access, unlimited exports, or the full product graph.
What it does
The Model Context Protocol (MCP) server exposes controlled intelligence tools: market summaries, buyer and supplier teasers, renewal signals, pursuit qualification, market change slices, and supplier comparisons. Every response is capped, confidence-labelled, and includes a link back to TenderLedger for deeper workflows.
Example tools
search_procurement_market— high-level slice for a topic (e.g. NHS cyber security).get_buyer_intelligence— compact buyer profile from indexed notices.get_supplier_intelligence— competitor-style profile from disclosed winners.find_likely_renewals— re-tender teasers from TenderLedger precompute.qualify_opportunity— pursuit / bid qualification (precomputed when a tender id is known).get_market_changes— what moved in the last N days for a topic slice.compare_suppliers— side-by-side comparison for up to three names.
Example prompts
- “Which suppliers are winning MOD-linked IT awards in the TenderLedger index?”
- “Summarise fire safety renewal signals for the next 12 months.”
- “Compare Softcat and Capita on disclosed public-sector awards.”
Safety limits
Default anonymous quotas apply per IP; API keys unlock higher daily limits. Result depth is capped (typically five items per dimension). There is no arbitrary SQL, no raw table dumps, and no access to other users’ private workspaces.
How to run locally
See backend/MCP.md in the repository for environment variables, stdio setup (Claude / Cursor), REST POST /api/mcp/v1/invoke, and streamable HTTP mounting.
Request access
MCP is gated by configuration on the API host. For production keys, higher quotas, or enterprise audit requirements, contact us from Book a demo or your TenderLedger account team.