Server data from the Official MCP Registry
Pay-per-call JSON repair + JSON Schema validation for AI agents (USDC on Base, x402).
About
Pay-per-call JSON repair + JSON Schema validation for AI agents (USDC on Base, x402).
Remote endpoints: streamable-http: https://x402.agentfund.net/mcp
Security Report
This is a well-architected pay-per-call MCP server with solid security practices. The codebase properly isolates payment logic from tool logic, refuses private keys in configuration, and uses environment variables for secrets. Code quality is good with proper input validation via Zod schemas and comprehensive test coverage. No critical vulnerabilities or malicious patterns detected. Minor code quality observations (broad exception handling, some defensive checks) do not materially affect the security posture. Supply chain analysis found 11 known vulnerabilities in dependencies (2 critical, 2 high severity).
5 files analyzed · 16 issues found
Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.
Permissions Required
This plugin requests these system permissions. Most are normal for its category.
How to Install & Connect
Available as Local & Remote
This plugin can run on your machine or connect to a hosted endpoint. during install.
Documentation
View on GitHubFrom the project's GitHub README.
US Economic, SEC EDGAR & On-Chain Data — a pay-per-call MCP server
21 data tools that AI agents discover and pay for per call, settled in USDC on Base via the x402 payment protocol. No account, no API key, no subscription.
Live: https://x402.agentfund.net · Agent guide · Verification ledger · Settlements
| Endpoint | Purpose |
|---|---|
POST /mcp | MCP over streamable-HTTP (stateless) |
POST /x402/<tool> | One HTTP route per tool; arguments as a plain JSON body |
GET /openapi.json | OpenAPI 3.1 discovery document with x-payment-info per route |
GET /SKILL.md | Agent-facing usage guide |
GET /VERIFICATION.md | How each tool was verified, and the upstream traps found |
GET /monitor | Live on-chain settlement dashboard |
GET /health | Liveness and configuration mode |
The tools
US macro — treasury_yield_curve, bls_cpi, macro_pce, macro_jobs, macro_gdp,
macro_retail_sales, macro_housing, macro_energy, macro_release_calendar
SEC EDGAR — edgar_filings_feed, edgar_financials, edgar_insider_transactions,
edgar_13f_holdings, edgar_full_text_search
On-chain EVM — onchain_token_balances, onchain_portfolio, onchain_cross_chain_balances,
onchain_oracle_price, onchain_gas
Pure compute — structured_json_repair, tabular_to_json
Prices run $0.001–$0.03 per call.
Why these sources
Every tool wraps free public-domain US government data (Treasury, BLS, BEA, Census, EIA, SEC EDGAR) or a direct on-chain read via public RPC. There is no upstream vendor licence, so nothing here can be revoked or repriced by a third party, and the cost of goods is zero.
The trade-off is inherited from the publishers: government statistics are lagged and revised, 13F is quarterly and stale by design, and none of this is market data. Those limits are stated plainly in SKILL.md rather than buried.
Correctness
Every tool was checked against live upstream data before shipping — not only against unit tests, because a test written from the same wrong assumption as the code passes happily.
That caught six real defects, including 13F values being 1000× too large by following SEC's own documentation, and Census silently returning five regional rows where a national figure was expected. Each is documented with the evidence that exposed it, and locked in by a regression test: VERIFICATION.md.
A failed call is never billed. The MCP SDK turns a thrown tool error into an HTTP 200 carrying
isError: true; the payment gate inspects the result and returns errors unsettled rather than
charging for a result the caller never received.
Architecture
- x402 v2 (
@x402/core,@x402/evm) with the Coinbase CDP facilitator on Base mainnet. - MCP on Workers via
@hono/mcpStreamableHTTPTransport(the SDK's own transport is Node-http-based). Stateless JSON. - Two entry points, one payment path.
/mcpand/x402/<tool>both run throughPaymentGate.chargeAndRun, so money-safety rules cannot drift between them. - Per-tool HTTP routes exist for discovery. The x402 Bazaar indexes plain HTTP resources only
— every catalog entry is
type: "http"— so an MCP endpoint alone can never be listed. - Discovery needs no secrets.
initialize,tools/listandpingare answered without payment configuration, so the server introspects cleanly on a fresh clone or in a CI sandbox. - Validation uses
@cfworker/json-schema, not Ajv — Ajv compiles viaFunction, which Workers forbid, and the schema is a runtime input so it cannot be precompiled. - No private keys. The server holds only a public payout address;
config.tsrefuses private-key- or seed-shaped input.
Adding a tool is a one-file change plus a line in src/tools/index.ts.
Development
npm install
npm test # 174 tests
npm run typecheck
npm run dev # wrangler dev
npm run dev:node # tsx watch src/node.ts
Copy .dev.vars.example to .dev.vars for local configuration. Only the public payout
address is ever needed — never a private key or seed phrase.
Docs are mirrored into the bundle (Workers have no filesystem); run npm run gen:docs after
editing SKILL.md or VERIFICATION.md, or test/docs.test.ts will fail on the drift.
Deployment
Production runs on Cloudflare Workers. Secrets are set with wrangler secret put, never in
wrangler.toml:
npx wrangler deploy --env production
A Dockerfile is included for self-hosting the Node entrypoint; it needs no secrets, so the
container starts and passes introspection out of the box. See
OPERATOR_CHECKLIST.md for the full runbook.
Buyer-side test clients
PAYER_PRIVATE_KEY=0x… node scripts/pay-test.mjs # pay via /mcp
PAYER_PRIVATE_KEY=0x… node scripts/pay-http.mjs # pay via /x402/<tool>
node scripts/index-all.mjs # dry run: Bazaar index status
Payment is gasless for the payer (the facilitator submits the transaction), so a test wallet
needs USDC and no ETH. index-all.mjs skips already-indexed routes and refuses to spend without
--yes.
Licence
MIT (declared in package.json; no LICENSE file has been added yet).
Reviews
No reviews yet
Be the first to review this server!
More Developer Tools MCP Servers
Fetch
Freeby Modelcontextprotocol · Developer Tools
Web content fetching and conversion for efficient LLM usage
Git
Freeby Modelcontextprotocol · Developer Tools
Read, search, and manipulate Git repositories programmatically
Toleno
Freeby Toleno · Developer Tools
Toleno Network MCP Server — Manage your Toleno mining account with Claude AI using natural language.
mcp-creator-python
Freeby mcp-marketplace · Developer Tools
Create, build, and publish Python MCP servers to PyPI — conversationally.
MCP Marketplace
Freeby mcp-marketplace · Developer Tools
Search and install MCP servers from inside your AI client.
MarkItDown
Freeby Microsoft · Content & Media
Convert files (PDF, Word, Excel, images, audio) to Markdown for LLM consumption
