Constat is an FDA and NHTSA regulatory risk intelligence MCP server for AI agents, providing real-time public regulatory data externally based on the Model Context Protocol (MCP). It is aimed at agent applications that need compliance risk information for medical devices, drugs, or vehicle safety, helping them obtain public data such as recalls, adverse events, warning letters, Form 483s, 510(k) premarket evidence, postmarket drift signals, reimbursement pathways, and vehicle safety in automated workflows.
Main capabilities: The service provides a total of 14 tools, covering topics such as recalls, adverse events, warning letters, 483s, 510(k) predicate chains, reimbursement, and vehicle safety. Each tool declares an output schema, and tool calls return structured content with explicit ok, not_found, invalid_request, or unavailable statuses, while retaining text results for older clients. The tools are observational, non-destructive operations, but because calls write limited quota and operational telemetry state, they declare readOnlyHint: false; they do not publish or modify FDA, CMS, NHTSA, or other third-party records.
Usage: The endpoint is https://constat.dev/api/mcp，旧地址 https://radar.healthai.com/api/mcp can still continue to serve. The transport method is Streamable HTTP (JSON-RPC 2.0). Anonymous access does not require authentication, but each anonymous source is limited to 10 work units per minute and 50 per day; an optional X-API-Key request header can enable separately allocated metered quota. Tool calls have documented work unit costs. The server's identifier in the official MCP registry is com.healthai/radar. The evidence corpus and tool logic are hosted and run at the above endpoint. This repository also provides server.mjs, which is a zero-dependency stdio bridge to that endpoint, enabling MCP clients that only support stdio to use the service like any local server. Use the tools/list method to list tools; note that the Accept request header must include text/event-stream (this is a requirement of Streamable HTTP).
Applicable scenarios: Suitable for developers building AI agents who need to integrate regulatory risk data, such as medical device compliance risk assessment, 510(k) premarket evidence and predicate chain queries, postmarket signal monitoring, reimbursement pathway analysis, and retrieval of vehicle safety-related regulatory information.
Limitations and notes: The service provides information and decision support only and does not constitute regulatory advice; when using it, verification against the original FDA/NHTSA sources should be performed. Anonymous access is subject to work unit quota limits. The specific tool parameters, details of returned fields, and the specific values of the metered quota are not provided in the materials. The project uses the MIT license.
