About Gao Protests
Gao Protests is an MCP server for researchers examining US government contract bid protests. It searches accumulated GAO decisions by outcome, company, or B-number, and reports that its coverage is not the complete historical archive.
Key features
- 01Filter recent protests by outcome
- 02Retrieve a decision and PDF link by B-number
- 03Search decisions by protester or company
- 04Check dataset coverage and crawl time
Use cases
- 01Review recent government contract disputes
- 02Find protests filed by a specific company
- 03Track the time span covered by the dataset
MCP support
It exposes tools for recent decisions, individual decisions, protester searches, and coverage details to MCP clients. It can connect through the Pipeworx gateway or run as a local stdio MCP server.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://gateway.pipeworx.io/gao-protests/mcpclaude mcp add --transport http pipeworx-io-gao-protests https://gateway.pipeworx.io/gao-protests/mcp{"mcpServers":{"pipeworx-io-gao-protests":{"type":"http","url":"https://gateway.pipeworx.io/gao-protests/mcp"}}}Open in Cursor{"servers":{"pipeworx-io-gao-protests":{"type":"http","url":"https://gateway.pipeworx.io/gao-protests/mcp"}}}Open in VS Code{"mcpServers":{"pipeworx-io-gao-protests":{"type":"http","url":"https://gateway.pipeworx.io/gao-protests/mcp"}}}{"mcpServers":{"pipeworx-io-gao-protests":{"serverUrl":"https://gateway.pipeworx.io/gao-protests/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project README@pipeworx/gao-protests
GAO bid-protest decisions — government-contract award disputes ruled on by the
US Government Accountability Office ("did anyone protest this award", "how
often does this agency lose protests"). Pairs with usaspending and
dod-contract-announcements, which cover the award side but not disputes.
Part of Pipeworx — an MCP gateway connecting AI agents to 1879+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.
Tools
gao_protests_recent(days?, outcome?, limit?)— recently decided protests, optionally filtered by outcome (denied|dismissed|sustained).gao_protest(b_number)— a single decision by B-number, with the decision PDF link.gao_protests_search(protester, limit?)— search accumulated decisions by protester/company name.gao_protests_coverage()— total decisions held, earliest/most-recent decided_on, last crawl time.
Auth
Keyless — no operator key needed. Backed by gao_bid_protests, refreshed
daily by workers/data-pipeline (dataset gao-protests).
Data sources
- https://www.gao.gov/legal/bid-protests/search — the bid-protests landing page. It embeds a Drupal Views "recent decisions" block directly in the page HTML (no JS challenge), showing the ~5 most-recent decisions.
- https://www.gao.gov/products/B-XXXXXX — each decision's own page, fetched once per new B-number for its exact decided date and PDF link.
Coverage is a rolling accumulation, not GAO's full protest archive. The
filtered search endpoint (gao.gov/search?f[0]=ctype_search:Bid Protest Decision) sits behind an Akamai Bot Manager proof-of-work JS interstitial
that cannot be solved from a server — full Chrome headers over HTTP/2 don't
help, this is a JS-execution challenge, not a header fingerprint. The landing
page's own "More" link points at that same blocked endpoint, and there is no
other reachable pagination. So this pack can only ever see the decisions GAO
has published since it started crawling — gao_protests_recent/_search
answer against what has accumulated in gao_bid_protests so far, not against
GAO's complete decision history. gao_protests_coverage reports exactly how
far back that goes.
Both hosts need the full Chrome header set (User-Agent, Accept,
Accept-Language, sec-ch-ua*, Sec-Fetch-*, Upgrade-Insecure-Requests
over HTTP/2) — a plain User-Agent alone gets a 403. This is a
header-fingerprint block, not a Cloudflare-egress block; no proxy is needed.
The protested AGENCY is not available. It appears only inside each decision's PDF text, never in the page HTML, and this pipeline has no PDF text extraction. Rather than guess an agency from context (or from the protester's history), the field is omitted entirely — presenting a guessed agency as fact would be worse than not having it.
A decision can carry several B-numbers (e.g. B-424484,B-424484.2 for an
original protest plus a reconsideration). The table's primary key is the
first number in that string; gao_protest also matches a non-primary number
(like the .2 suffix) by searching the raw string.
Outcome is prose, not an enum ("We deny the protest.", "We dismiss the
request for reconsideration.", "We sustain the protest."). It's normalized to
denied | dismissed | sustained | other for filtering, and the raw sentence
is always kept alongside it.
Quick Start
Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):
{
"mcpServers": {
"gao-protests": {
"url": "https://gateway.pipeworx.io/gao-protests/mcp"
}
}
}
What this endpoint actually serves
tools/list at https://gateway.pipeworx.io/gao-protests/mcp returns the tools in the table
above plus the shared Pipeworx meta-tools — ask_pipeworx,
discover_tools, search_within, remember/recall and the rest of the
gateway-wide set. So the tool count you see is larger than this table: a
single-pack endpoint currently lists roughly 30 shared tools alongside the
pack's own. The connection's initialize response states its exact scope, and
is the authoritative answer for a given day.
This is deliberate, not multiplexing by accident. The meta-tools are what let a
scoped connection answer a question this pack does not cover — via
ask_pipeworx, which routes across the whole catalog — without you adding a
second MCP server. There is currently no way to mount a pack endpoint without
them; if the extra schemas cost you more context than the routing is worth,
connect to the full gateway once rather than to several pack endpoints.
Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:
{
"mcpServers": {
"pipeworx": {
"url": "https://gateway.pipeworx.io/mcp"
}
}
}
Both URLs reach the same gateway and the same 1879+ data sources. The
only difference is which pack's tools are listed directly; ask_pipeworx
reaches all of them from either one.
No MCP client? Call it over HTTP
curl -X POST https://gateway.pipeworx.io/v1/tools/gao_protests_recent \
-H 'Content-Type: application/json' \
-d '{"days":30}'
No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/gao_protests_recent. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.
Standalone (no gateway account)
This package also runs as a local stdio MCP server — no Pipeworx account, no gateway round-trip:
{
"mcpServers": {
"gao-protests": {
"command": "npx",
"args": ["-y", "@pipeworx/mcp-gao-protests"]
}
}
}
Or run it directly to confirm it starts:
npx -y @pipeworx/mcp-gao-protests
It speaks MCP over stdin/stdout and answers initialize/tools/list/tools/call
for only this pack's tools — none of the shared meta-tools the gateway
connection above adds. Same source, same tools, no ask_pipeworx routing.
Using with ask_pipeworx
Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:
ask_pipeworx({ question: "your question about Gao Protests data" })
The gateway picks the right tool and fills the arguments automatically.
More
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.