Fourteen tools for querying FDA data across drugs, food, devices, animal/veterinary products, and recalls — plus an optional DataCanvas SQL surface for large result sets:

Resolve one drug name to its FDA identity, then return a consolidated profile in a single call — replacing four or five chained lookups.

• Resolves a brand or generic name to canonical FDA identifiers once (generic name, NDC, RxCUI, SPL set ID), then keys every sub-query off that identity to avoid the identifier drift that breaks naive tool chaining
• Single-ingredient resolution: a single-drug query won't resolve to a combination product
• Sections: label highlights, adverse-event summary (top reactions, serious count), recall history, Drugs@FDA approval, and current shortage status
• Best-effort — a miss on any section returns null rather than failing the whole call; use the dedicated tool for a deep dive into any area

Search adverse event reports across drugs, food, and devices. Use to investigate safety signals, find reports for a specific product, or explore reactions by demographics.

• Category selection: drug, food, or device — each returns different field schemas
• Elasticsearch query syntax for filtering by product, reaction, seriousness, date range
• Pagination via limit (up to 1000) and skip (up to 25000)
• Formatted output includes report ID, seriousness, patient demographics, reactions, drugs with characterization/indication/route, and all remaining fields

Aggregate and tally unique values for any field across any openFDA endpoint. Returns ranked term-count pairs sorted by count descending.

• Works across all 20 openFDA endpoints (drugs, food, devices, animal/veterinary, tobacco, other)
• Use .exact suffix on field names for whole-phrase counting
• Optional search filter to scope the aggregation
• Returns up to 1000 terms per query

Search enforcement reports and recall actions across drugs, food, and devices.

• Supports enforcement (all categories) and recall (devices only) endpoints
• Filter by classification (Class I/II/III), recalling firm, reason, status
• Formatted output includes recall number, classification, product description, reason, distribution pattern

Search FDA device premarket notifications — 510(k) clearances and PMA approvals.

• Two pathways: 510k (174K+ records, most common) and pma (higher-risk devices)
• Filter by applicant, product code, advisory committee, device name
• Formatted output adapts to pathway: 510(k) shows K-number/clearance type, PMA shows supplement info

Look up FDA drug labeling (package inserts / SPL documents). Check indications, warnings, dosage, contraindications, active ingredients, or any structured label section.

• Search by brand name, generic name, manufacturer, or set ID
• Formatted output dynamically renders all label sections and openfda metadata present in the record
• Large sections are automatically truncated to keep output readable
• Default limit of 5 — labels are large documents

Search the Drugs@FDA database for drug application approvals (NDAs and ANDAs). Returns application details, sponsor info, and full submission history.

• Filter by brand name, sponsor, submission type, review priority
• Formatted output includes products with active ingredients, dosage forms, routes, and marketing status
• Full submission history with type, status, date, and review priority
• Pagination via limit (up to 1000) and skip (up to 25000)

Look up drugs in the NDC (National Drug Code) Directory. Identify drug products by NDC code, find active ingredients, packaging details, or manufacturer info.

• Search by product NDC, brand name, generic name, manufacturer, or active ingredient
• Returns product details, active ingredients with strengths, and packaging information
• Sortable by listing expiration date or other fields

Search adverse event reports for veterinary drugs and devices submitted to the FDA Center for Veterinary Medicine (1.3M+ records).

• Filter by animal species, breed, drug name, VeDDRA reaction term, or seriousness
• Records include animal details (species, gender, age, weight), administered drugs, reactions, and outcomes
• Formatted output surfaces key clinical fields; remaining fields rendered via catch-all

Search problem reports submitted to the FDA for tobacco products, including e-cigarettes, vaping products, cigarettes, and smokeless tobacco.

• Filter by product type, reported health problems (e.g. seizure, chest pain), product problems (e.g. battery explosion), or non-user involvement
• Formatted output surfaces products, health effects, product defects, and report counts

Search FDA drug shortage records (1,700+ entries, refreshed daily). Returns shortage status, availability notes, therapeutic category, dosage form, manufacturer, and timeline.

• Filter by status (Current, Resolved), therapeutic category, generic name, or manufacturer
• The openfda block carries brand_name, product_ndc, and rxcui for chaining into openfda_get_drug_label or openfda_lookup_ndc
• Pagination via limit (up to 1000) and skip (up to 25000)

Return the searchable field paths for an openFDA endpoint, grouped by category with type and description. Use before constructing a search query to discover the correct dotted field paths.

• Covers all major endpoints: drug/event, drug/label, drug/shortages, drug/drugsfda, drug/ndc, drug/enforcement, food/event, food/enforcement, device/event, device/510k, device/pma, device/recall, device/enforcement, animalandveterinary/event, tobacco/problem
• Returns fields grouped by category (identifiers, dates, clinical fields, etc.) with data type and one-line description
• Complements the reactive field hints that appear in notice enrichment when a search returns empty

A DataCanvas SQL surface over staged result sets — opt-in, enabled with CANVAS_PROVIDER_TYPE=duckdb.

• With canvas enabled, the multi-row search tools page the full matched set (up to openFDA's 25,000-row ceiling) into a DuckDB table and return an inline preview plus a canvas_id and canvas_table. openfda_dataframe_query runs read-only SELECT (GROUP BY, SUM/COUNT, joins) across the staged rows; openfda_dataframe_describe lists the table and column schemas needed to write valid SQL.
• Scalar fields are stored as text (CAST for numeric math); nested openFDA blocks (openfda, patient, products, …) are JSON columns. Pass a canvas_id back into a search tool to accumulate result sets on one canvas for cross-table joins.
• Off by default — without CANVAS_PROVIDER_TYPE=duckdb the search tools return inline results exactly as before and the two dataframe tools report that canvas is disabled. Requires the optional @duckdb/node-api dependency; unsupported on Cloudflare Workers.

Built on @cyanheads/mcp-ts-core:

• Declarative tool definitions — single file per tool, framework handles registration and validation
• Unified error handling across all tools
• Pluggable auth (none, jwt, oauth)
• Swappable storage backends: in-memory, filesystem, Supabase, Cloudflare KV/R2/D1
• Structured logging with optional OpenTelemetry tracing
• Runs locally (stdio/HTTP) or on Cloudflare Workers from the same codebase

openFDA-specific:

• Generic API client for all openFDA endpoints with retry (exponential backoff) and rate-limit awareness
• Automatic error normalization — 404 returns empty results, 429/5xx retries, 400 provides actionable messages
• Optional API key support — works without a key (1K requests/day), increases to 120K/day with a free key
• Optional DataCanvas spillover (CANVAS_PROVIDER_TYPE=duckdb) — stage large result sets as DuckDB tables and run SQL via openfda_dataframe_query

A public instance is available at https://openfda.caseyjhand.com/mcp — no installation required. Point any MCP client at it via Streamable HTTP:

{
  "mcpServers": {
    "openfda-mcp-server": {
      "type": "streamable-http",
      "url": "https://openfda.caseyjhand.com/mcp"
    }
  }
}

Add to your MCP client config:

{
  "mcpServers": {
    "openfda-mcp-server": {
      "type": "stdio",
      "command": "bunx",
      "args": ["@cyanheads/openfda-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info",
        "OPENFDA_API_KEY": "your-key-here"
      }
    }
  }
}

Or with npx (no Bun required):

{
  "mcpServers": {
    "openfda-mcp-server": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cyanheads/openfda-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info",
        "OPENFDA_API_KEY": "your-key-here"
      }
    }
  }
}

Or with Docker:

{
  "mcpServers": {
    "openfda-mcp-server": {
      "type": "stdio",
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "MCP_TRANSPORT_TYPE=stdio", "ghcr.io/cyanheads/openfda-mcp-server:latest"]
    }
  }
}

For Streamable HTTP, set the transport and start the server:

MCP_TRANSPORT_TYPE=http MCP_HTTP_PORT=3010 bun run start:http
# Server listens at http://localhost:3010/mcp

• Bun v1.3.0 or higher.
• Optional: openFDA API key for higher rate limits (120K requests/day vs 1K/day).

1. Clone the repository:

git clone https://github.com/cyanheads/openfda-mcp-server.git

2. Navigate into the directory:

cd openfda-mcp-server

3. Install dependencies:

bun install

All configuration is validated at startup via Zod schemas in src/config/server-config.ts. Key environment variables:

• Build and run the production version:

  # One-time build
  bun run rebuild
  
  # Run the built server
  bun run start:http
  # or
  bun run start:stdio

• Run checks and tests:

  bun run devcheck  # Lints, formats, type-checks, and more
  bun run test      # Runs the test suite

See CLAUDE.md for development guidelines and architectural rules. The short version:

• Handlers throw, framework catches — no try/catch in tool logic
• Use ctx.log for request-scoped logging
• Register new tools in src/mcp-server/tools/definitions/index.ts

Issues and pull requests are welcome. Run checks and tests before submitting:

bun run devcheck
bun run test

This project is licensed under the Apache 2.0 License. See the LICENSE file for details.