for developers & AI agents

Use Argus HQ inside Claude, ChatGPT, and Perplexity

Argus HQ's FDA enforcement database is fully machine-readable: an MCP server, four per-entity JSON APIs, an OpenAPI 3.1 spec, and an llms.txt file. Every surface below is free and requires no API key.

1. MCP server

Argus HQ runs a Model Context Protocol (MCP) server at /api/mcp exposing five tools: search_enforcement, get_company_history, get_recall_by_lot, get_warning_letter, and list_recent_actions. No auth required — this data is already fully public.

{
  "mcpServers": {
    "argus-hq": {
      "url": "https://argushq.ai/api/mcp"
    }
  }
}

Claude Desktop

Settings, Developer, Edit Config -- paste the JSON above into the mcpServers block, then restart Claude Desktop.

ChatGPT (connectors / GPT Actions)

ChatGPT connectors currently prefer OpenAPI-based Actions over raw MCP -- use the OpenAPI spec in Section 3 below to add Argus HQ as a custom GPT Action.

Cursor

Cursor Settings, MCP, Add new MCP server -- paste the same JSON config (Cursor uses the identical mcpServers schema).

Windsurf

Windsurf Settings, Cascade, MCP Servers, Add server -- same JSON config as Claude Desktop.

Listed on

Namespace ai.argushq/* is DNS-verified against argushq.ai.

Official MCP Registry →

2. JSON APIs

Four per-entity JSON endpoints, no API key, cacheable, CORS-open. Each returns a single record as structured JSON with the same fields documented in the field dictionary on the data-licensing page.

Warning letter

/api/warning-letter/{slug}.json

curl https://argushq.ai/api/warning-letter/example-co-2026-01-15.json

Recall

/api/recall/{slug}.json

curl https://argushq.ai/api/recall/example-co-recall-2026-02-01.json

Approval

/api/approval/{slug}.json

curl https://argushq.ai/api/approval/example-drug-approval-2026.json

Company rollup

/api/company/{slug}.json

curl https://argushq.ai/api/company/pfizer-inc.json

3. OpenAPI spec

A full OpenAPI 3.1 spec documents every public endpoint above plus the RSS feeds. Point GPT Actions, n8n, Zapier, or any OpenAPI-aware tool at it directly.

Download openapi.json

Adding as a Custom GPT Action

In ChatGPT, create or edit a Custom GPT.
Under Configure, Actions, Create new action.
Choose Import from URL and paste https://argushq.ai/openapi.json.
Authentication: None -- every endpoint is public.
Save. Your GPT can now query Argus HQ FDA enforcement records directly.

4. llms.txt and machine-readable feeds

A plain-text llms.txt file (the emerging llmstxt.org convention) describes the whole site for AI crawlers and assistants, plus bulk JSON and RSS feeds for anything that prefers polling over MCP.

/llms.txt -- plain-text site description for LLMs
/data/enforcement-index.json -- full bulk JSON index
/feed/warning-letters.xml -- RSS, Warning Letters
/feed/recalls.xml -- RSS, Recalls
/feed/approvals.xml -- RSS, Approvals

5. Dataset access

For bulk licensing, sample downloads, a field dictionary, and live record counts, see /data-licensing.

GitHub mirror -- coming soon.
Kaggle mirror -- pending.

Questions, rate-limit needs, or a use case that doesn't fit the surfaces above? Email hello@argushq.ai.