What it does
This server runs the agent-tools.cloud directory of x402 paid APIs, MCP servers and A2A agents. Services are probed daily, graded and safety-scanned. Listings, daily statistics, the scoring methodology and blog articles are available as JSON and through MCP tools. A free MCP discovery endpoint offers search, a ranked recommendation tool (ask_services) and a safety scan (scan_mcp_safety). An A2A agent on /a2a provides directory search, MCP safety scanning and daily statistics. A core subset of the search tools is also published as a standalone PyPI package.
It suits finding a service by intent, such as searching in natural language and reviewing match reasons and confidence scores. Before connecting to an MCP server, the safety scan checks its advertised name, description and tool documentation with deterministic static rules that do not execute code. The rules flag pipe-to-shell installs, bare-IP payload hosts and tool-poisoning phrasing. A frontier LLM adds an advisory second opinion. Paid verification lets a user pay a listed x402 service once from their own wallet and record the result.
The service runs behind nginx as systemd units, with two uvicorn instances deployed by rolling restart. Daily crawl and health-check timers are included. Configuration is set through an .env file, covering upstream and LLM backend credentials, the SQLite database path and rate limits for the public ask endpoint. Articles are disabled by default and must be enabled with an environment flag.
Limitations and requirements
Paid verification requires an API key from POST /api/v1/keys and is available only on the remote server, not in the PyPI package, which ships only the core search tools. The upstream model listing and the ask endpoint need their own API keys and LLM backend settings. The public ask endpoint is rate-limited. The scan does not apply source-code injection rules to natural-language descriptions, to avoid false positives. The paid relay and paid vertical endpoints were retired on 2026-05-25.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 14 Nov 2026 · its page on the source
agent-tools (x402 + MCP + A2A directory)
Discover x402 services, MCP servers and A2A agents by intent — agent-tools.cloud directory.
v0.5.2As shown on registry.modelcontextprotocol.io when we last read it