MCP vs REST API
Two ways to access the same data. Choose the one that fits your use case.
MCP (Model Context Protocol)
Best for AI agents, coding assistants, and conversational tools
- How it works: Your AI calls tools directly --
search,get_document,resolve_reference, and more. No HTTP code needed. - Auth: OAuth -- automatic sign-in on first use. Alternatively, pass an API key in headers.
- Advantages: Structured tool calls, automatic capability discovery, rich context for the model.
- Setup: Paste
https://legaldatahunter.com/mcpinto your MCP client (Claude, Cursor, ChatGPT, Windsurf, etc.).
REST API
Best for web apps, mobile apps, dashboards, scripts, and data pipelines
- How it works: Standard HTTP requests to
/v1/*endpoints. JSON in, JSON out. - Auth: Bearer token -- add
Authorization: Bearer sk-...to every request. - Advantages: Full control, works everywhere HTTP works, easy to integrate into any stack.
- Setup: Get a key at /account, then start making requests.
Side-by-Side Comparison
| MCP | REST API | |
|---|---|---|
| Best for | AI agents & assistants | Apps, UIs, scripts |
| Protocol | Streamable HTTP (MCP) | Standard HTTP / JSON |
| Auth | OAuth (automatic) | Bearer token |
| Tool discovery | Automatic | Docs or /swagger |
| Data access | Same corpus | Same corpus |
| Rate limits | Same tiers | Same tiers |
| Latency | Same | Same |
Which Should I Use?
- Building an AI agent or using an AI assistant? → MCP
- Building a web app, mobile app, or dashboard? → REST API
- Writing a data pipeline or backend script? → REST API
- Want both? → Use both!
MCP and REST share the same authentication and the same data. You can use both in the same project -- for example, MCP for your AI features and REST for your dashboard.
Both Use the Same Backend
MCP tools map 1:1 to REST endpoints. The search tool calls POST /v1/search, get_document calls GET /v1/documents/{source}/{source_id}, and so on. Same rate limits, same data, same response format.
The MCP server acts as a thin adapter that translates tool calls into API requests. There is no separate data path or caching layer -- every MCP call hits the same infrastructure as a direct API request.
MCP Tool to REST Endpoint Mapping
MCP Tool REST Endpoint search POST /v1/search get_document GET /v1/documents/{source}/{source_id} resolve_reference POST /v1/resolve discover_countries GET /v1/discover/countries discover_sources GET /v1/discover/countries/{code}/sources get_filters GET /v1/discover/sources/{source}/filters report_source_issue POST /v1/sources/{source}/issues