INTEGRATION · Pro Max

Use BidSparq from any AI chat.

Connect Claude, Cursor, ChatGPT, or any MCP-compatible client. Your AI assistant gets instant access to 75 tools across 35K active RFPs, 1.6M closed RFPs, 164K federal awards, 102K subaward records, and 19K SAM-registered vendors.

MCP server URL

https://bidsparq.com/mcp
OAuth 2.1 + Bearer keyStreamable HTTP transportListed in the official MCP Registry
Connect Claude Desktop →See what you can ask

Works with the MCP-compatible AI clients you already use

Claude Desktop·Claude Code·Cursor·ChatGPT·Microsoft Copilot·Cline·RooCode·Continue

How it works

Ask in plain English. Get back real data.

Once connected, your AI client calls BidSparq tools the same way it calls its built-in ones. Here's what you can ask — and what comes back.

  • “Find me primes who hire subs in NAICS 541512.”

    find_subaward_primes
    Top 20 primes ranked by sub-hiring volume — SAIC, GDIT, Peraton, Booz Allen — with sub_count, total_subawards in dollars, top awarding agencies, and example subawards. SAM-matched UEIs deduped across acquisitions (Perspecta + Peraton).

  • “What's the typical award amount for VA software contracts?”

    analyze_pricing_intel
    p25 / median / p75 across 2,147 matching awards, plus year-over-year trend and per-agency average. Real percentile values straight from USAspending — no estimates.

  • “Show me wired RFPs in defense IT so I know what to skip.”

    filter_rfps (wired_only=true)
    RFPs flagged by our wired-signal detector: brand-name lock-in, intent-to-award language, thin PWS, competitor-specific wording — with sourced excerpts.

  • “What contracts come up for recompete in cybersecurity over the next 12 months?”

    find_recompetes
    Expiring contracts with the incumbent, days-until-expiry, contract value, and matching active RFPs by NAICS so you can see the new vehicle alongside the legacy one.

  • “Tell me about NASA's buying patterns.”

    get_agency_profile
    Top NAICS codes, top vendors by spend, seasonal pattern (Q4 surge), max award size, competition density, and total annual obligated spend. Sentinel/placeholder values filtered out.

  • “Find HUBZone businesses in Texas for NAICS 541512.”

    search_vendors
    SAM-registered vendors with NAICS match + HUBZone certification active, including UEI, certifications, primary NAICS, location, and registration status.

Setup · under 2 minutes

Three steps and you're connected.

  1. 1

    Upgrade to Pro Max

    MCP access is included on the Pro Max plan — no separate fee. Unlimited usage subject to fair-use limits, shared with the in-app AI chat. See pricing →

  2. 2

    Paste the URL in your AI client

    Open your client's MCP settings, paste https://bidsparq.com/mcp. Specific instructions per client below.

  3. 3

    Click “Allow” in your browser

    Your AI client opens an OAuth consent page on BidSparq. Click Allow once — done. The connection auto-refreshes for 30 days. Revoke anytime at Settings → Connected Apps.

Setup in your AI client

Pick your client. Each has slightly different settings UI; the URL is the same.

Claude Desktop's built-in custom-connector UI handles the OAuth flow for you — no JSON editing needed.

  1. 1Open Claude Desktop and go to Settings → Connectors.
  2. 2Click Add custom connector.
  3. 3Paste https://bidsparq.com/mcp as the URL. Leave headers blank — Claude will start the OAuth dance automatically.
  4. 4Click Connect. Your browser opens to BidSparq's consent page. Sign in if needed, then click Allow.
  5. 5Back in Claude Desktop, the BidSparq tools appear in the + menu of any chat. Try: “Find primes who hire subs in NAICS 541512”

Capabilities

All 75 tools, grouped by what you'd use them for.

RFP discovery

Find what to bid on next.

  • semantic_search

    Natural-language topic search across 35K active RFPs (pgvector HNSW + OpenAI embeddings).

  • filter_rfps

    Structured filter: state, agency, NAICS prefix, set-aside, due dates, wired-only, keywords, sort by value.

  • count_rfps

    Aggregate counts grouped by state / agency / NAICS. Filters out scraper-placeholder agencies.

  • list_sources

    All 130+ procurement sources we scrape with active RFP counts.

  • search_historical_bids

    1.6M+ closed RFPs with close-date sentinel cleanup — the deepest historical corpus in the market.

  • upcoming_deadlines

    Active RFPs closing in the next N days (default 7). Filterable by NAICS / agency / state / set-aside-only. Sorted most-urgent first.

RFP analysis

Decide if an RFP is worth bidding.

  • get_rfp_details

    Full enrichment for one RFP: AI summary, win-fit, complexity, compliance items, wired signals, related awards.

  • get_compliance_matrix

    Extracted compliance requirements with section + page references straight from the PDF.

  • compare_rfps

    Side-by-side comparison of 2-5 RFPs — value, NAICS, set-aside, AI win-probability, wired risk. Surfaces unresolved IDs.

  • search_rfp_documents

    Full-text + semantic search inside attached PDFs scoped to one RFP.

Market intelligence

The BidSparq-only tools — built on our 1.6M historical RFPs, 164K awards, 102K subawards, 19K SAM entities.

  • find_subaward_primes

    Primes who hired subs in your NAICS — 102K FFATA records, 18-month window. UEI-deduped across acquisitions.

  • vendor_win_history

    Per-vendor federal win history from 164K awards. By UEI or fuzzy name. Returns per-award detail + top agencies + cadence.

  • competitive_landscape

    Win concentration (top-1/3/10 % share), top vendors, avg bidders, set-aside mix, competition extent. The 'who am I really up against?' tool.

  • set_aside_analysis

    Per-set-aside breakdown for a NAICS: award counts, avg bidders, easier-than-baseline flag, top vendor per category. Use to evaluate certification ROI.

  • analyze_pricing_intel

    p25 / median / p75 POST-AWARD amounts (what was actually obligated). Real percentiles from USAspending.

  • budget_intelligence

    p25 / median / p75 PRE-RFP POSTED budget ranges from 1.6M historical RFPs. Planning $ vs. obligated $.

  • contract_duration_intel

    Typical contract length from 1.6M closed RFPs. Parses '1 year' / '36 months' / '5 years' patterns into months percentiles + buckets.

  • search_awards

    Past contract awards by agency / NAICS / vendor. Filters $1T ceiling-amount sentinels.

  • find_recompetes

    Expiring contracts + incumbent + days-until-expiry + matching active RFPs (NAICS-locked).

  • get_agency_profile

    Spend stats, top vendors, top NAICS, max-award + competition density. VA → Veterans Affairs abbreviation routing.

  • get_seasonal_analysis

    Monthly award patterns by NAICS / agency — find when an agency actually buys.

  • get_trending

    Weekly new-RFP counts grouped by state / source / bid type.

  • search_vendors

    19K SAM-registered vendors with NAICS arrays, set-aside certs, UEI, location, expiration.

Personal relevance (signed-in only)

Uses BidSparq's 75K-row contractor×RFP scoring engine to answer 'what should I bid on?'

  • get_my_top_matches

    Your highest-AI-scored active RFPs. Filterable by min_score, state, NAICS, decision verdict. Returns ai_score + llm_pwin + bid_recommendation per match.

  • explain_my_fit

    Per-RFP fit explanation: ai_score, Shipley Pwin, factor breakdown (capability/past-perf/set-aside/geography/complexity), compliance gaps, wired-risk.

  • find_similar_active_rfps

    Query-by-example: given an RFP id, find other ACTIVE RFPs whose scope reads like it. Uses pgvector HNSW embedding similarity, not just NAICS/agency filters.

  • get_pipeline

    Your bid pipeline (new_match → reviewing → preparing → submitted → won/lost) with notes.

Security & privacy

You stay in control of what your AI client can do.

OAuth 2.1 with PKCE

Connection uses OAuth 2.1 — the same standard GitHub, Google, and your bank use. PKCE protects against code interception. No client secret to leak.

One limited scope (mcp:tools)

Your AI client gets a single scope — mcp:tools — to search your BidSparq data and manage your own pursuits (e.g. save an RFP, set a next action, advance a stage). It cannot change your account settings, billing, password, or any other connected app.

Conversation history stays in your client

BidSparq only sees the tool calls and their arguments (e.g., NAICS code, agency name). The wider conversation never leaves your AI client.

Revoke in one click

At Settings → Connected Apps you can see every client connected and disconnect any of them. All tokens for that client die immediately.

Tokens are hashed at rest

Access + refresh tokens are stored as sha256 hashes only. Even our database backups don't contain reversible credentials.

Auto-rotation, single-use refresh

Access tokens expire in 1 hour. Refresh tokens rotate on every use. Reuse of an old refresh token triggers a full revocation of the connection.

Alternative

Prefer static API keys for scripts and CI?

OAuth is the recommended path for everyday AI-client use. For scripted / headless / CI use cases, you can also use a long-lived Bearer key. Generate one at Settings → API Keys and paste it into the Authorization header of any of the configs below.

Remote URL (recommended for keys too)

{
  "mcpServers": {
    "bidsparq": {
      "url": "https://bidsparq.com/mcp",
      "headers": {
        "Authorization": "Bearer bsq_live_..."
      }
    }
  }
}

npm package (for clients without remote-MCP support)

{
  "mcpServers": {
    "bidsparq": {
      "command": "npx",
      "args": ["-y", "bidsparq-mcp-server"],
      "env": {
        "BIDSPARQ_API_KEY": "bsq_live_..."
      }
    }
  }
}

Frequently asked

Is the MCP integration free?
Included with the Pro Max plan ($249/mo or $199 on annual). No additional usage fees. Unlimited usage subject to fair-use limits, whether you use the in-app chat or your own MCP client. See /learn/fair-use for details.
What's the rate limit?
50 tool calls / day per account (shared with in-app chat). Plus a 60 requests / minute soft cap to prevent runaway loops. If you regularly need more, email [email protected].
Can I use this for production / automation?
Yes. Use a static API key (Settings → API Keys) rather than OAuth for headless / CI workloads. The same tools, same rate limits, same data.
What happens if I downgrade from Pro Max?
Your existing OAuth connections immediately stop working — the /mcp endpoint enforces tier on every call. Upgrade back and they resume.
What if my AI client uses the wrong tool?
Be specific in your prompt — e.g., 'use find_subaward_primes' instead of 'find me partners'. The tool descriptions are tuned for Claude 3.5+ and GPT-4-class models; smaller models may need more guidance.
Can my organization audit what's been accessed?
Yes. Every tool call is logged with timestamp, tool name, and arguments. Email [email protected] for an audit export.
Do you sell or share my queries?
No. Tool-call arguments stay on BidSparq's servers (US-based) and are only used to serve your request and for our internal abuse-prevention logs. We never sell or share your usage with third parties.
What if the MCP server is down?
Your AI client gets a clear error message and can fall back to its own knowledge. Our uptime target is 99.9% — track real-time at bidsparq.com/status (coming soon).

Troubleshooting

Tools missing in Claude Desktop / Cursor after install

Fully quit and relaunch the app — MCP servers are only loaded at startup. In Claude Desktop, click the + icon in the prompt area; bidsparq should appear in the tools list.

401 Unauthorized — Invalid or missing credentials

If using OAuth: your token may have expired or been revoked. Re-run the connection from your AI client. If using a Bearer key: the key may have been revoked. Create a new one at Settings → API Keys.

403 Forbidden — Pro Max plan required

Your account is on Free or Pro. Upgrade at /plan. MCP access flips on within seconds of upgrade.

429 Daily message limit reached

Pro Max plans have 50 tool calls per day (shared with in-app chat). Resets at midnight UTC. If you're regularly hitting it, contact us — we're gathering usage data on heavy users.

`fetch failed` on first npx run

The npx command needs network access to fetch the package and call BidSparq's API. Make sure your firewall doesn't block outbound HTTPS to bidsparq.com and registry.npmjs.org.

Tool returns nothing for a specific NAICS

Try a prefix instead of an exact code — e.g., '5415' instead of '541512'. Some scrapers tag RFPs with broader categories. The naics field on most tools accepts both.

Ready to put your AI assistant to work?

Pro Max plan, then paste https://bidsparq.com/mcp into your AI client. Click Allow. Done.

Upgrade to Pro MaxManage connected apps

Questions? [email protected]