M
Model Context Protocol

Give your AI agent the federal contracting intelligence it's missing.

Mindy MCP is a hosted server that hands Claude, Cursor, ChatGPT, or any agent you build a catalog of grounded GovCon tools — so the answers come from real federal data, not the model's imagination.

Connect free — 100 creditsSee pricing

No API key — you sign in through your browser. No card.

Your agent is smart. On federal contracting, it's guessing.

Ask a general-purpose agent a GovCon question and it invents a plausible answer. The intelligence isn't in the model — it's in SAM.gov, USASpending, SEC EDGAR, GSA CALC, and eight years of proprietary GovCon work.

"Who’s the incumbent on this VA cybersecurity recompete, and when does it expire?"

The model invents a plausible company and a made-up date.

"What’s the fair GSA labor rate for a Senior Software Engineer?"

The model produces a number with no source.

"Who at the Army Corps LA District actually buys this?"

The model returns a generic contracting.officer@army.mil that doesn’t exist.

Mindy MCP is the bridge — grounded tools that return the real company, the real date, the real person who buys.

Wrapping SAM.gov is the price of entry. Not the moat.

SAM.gov and USASpending are free public APIs — anyone can wrap them in a weekend. Mindy leads with those because they're genuinely useful, but the reason an agent stays is the layer competitors cannot copy:

  • The winning-playbook corpus
    8 years of course, proposal-template & podcast content that answers "how do I actually win this" — no public API has it.
  • Office-level buying contacts
    The named contracting officers & small-business POCs at a specific buying office — not the whole-department firehose.
  • A curated SBLO teaming roster
    The Small Business Liaison at 200 primes, re-researched and verified — so an agent knows who to call to team.
  • The podcast lesson corpus
    Real lessons from real contractor & agency guests, matched by topic, agency, or set-aside.

53 tools, across four layers

From open-opportunity search to a full stateless proposal loop. A few from each layer — the full catalog and per-call pricing live on the pricing page.

01

Public data & search

The commodity layer — genuinely useful, and where most agents start. Open solicitations, forecasts, expiring recompetes, the vehicles work flows through.

  • search_sam_opportunitiesOpen solicitations by keyword / NAICS / set-aside
  • get_expiring_contractsContracts expiring soon — your recompete targets
  • get_agency_forecastsPlanned buys 6–18 months before they hit SAM
  • extract_statement_of_workThe SOW/PWS pulled out as clean text
02

Competitive intelligence

Who you’re up against and who to team with — the incumbent, the capable firms, the price-to-win, the teaming front door.

  • get_incumbent_financialsSEC-EDGAR financials of a public incumbent
  • get_pricing_intelGSA CALC price-to-win labor rates (p25/p50/p75)
  • find_capable_contractors"Who can actually win this" — capable-firm scan
  • get_sblo_contactThe Small Business Liaison at a prime — the teaming door
03

Agency & award intelligence

The buyer beneath the department label — the specific buying office, its budget trend, its set-aside behavior, and the named people who actually buy.

  • search_federal_contactsNamed POCs at a specific buying office (~167K rows)
  • get_sba_goaling_shareSmall-business goals vs. actual set-aside obligations
  • search_agency_opps_by_officeOpportunities anchored to one buying office
  • get_regulatory_demandFederal Register signals — demand before SAM
04

Proprietary & proposal

The moat — 8 years of GovCon coaching no public API holds, plus a full stateless bid loop: bid/no-bid → compliance matrix → outline → draft → independent referee.

  • get_winning_playbookGrounded "how to win this" coaching — the moat
  • evaluate_bid_decisionThe 5-gate / 10-factor bid/no-bid, scored
  • extract_compliance_matrixEvery shall/must + Section L/M/C requirement
  • referee_proposal_complianceAn independent model reviews the draft vs. the matrix

The no-fabrication contract

Grounding is the product. Every tool returns a machine-readable signal for whether the answer is real: groundedmeans at least one real record backs it; a genuine miss says so plainly instead of inventing one. An agent knows when to trust the answer — and Mindy never maps a record to a NAICS, set-aside, or financial the source doesn't carry.

Built for the agent you already use

Claude · Claude Code · Cursor · ChatGPT · Copilot · or your own. Add one endpoint, sign in through the browser, and your agent can run a real BD task — find the opportunity, size the market, vet the incumbent, draft the response.

Connect your agent — freeSee pricing

100 free credits on your first connect — no card required.