Skip to content
unlob
Get an API key

unlob for Cursor

Cursor supports MCP servers directly. The code vertical and the docs content type make it particularly effective for finding authoritative documentation rather than blog posts about documentation.

Config lives in .cursor/mcp.json

.cursor/mcp.jsonjson
{
  "mcpServers": {
    "unlob": {
      "command": "npx",
      "args": ["-y", "@unlob/mcp"],
      "env": { "UNLOB_API_KEY": "ulb_your_key_here" }
    }
  }
}

Setup

  1. Get an API key

    10,000 free requests a month.

  2. Add the config

    Create .cursor/mcp.json in your project with the block above, or configure it globally in Cursor Settings under MCP.

  3. Verify

    The MCP settings pane lists connected servers and their tools.

  4. Scope your searches

    Ask for documentation specifically. content_type=docs and vertical=code remove the tutorial blogspam that dominates most programming queries.

Worth knowing

  • vertical=code is disambiguation by construction — a search for "rust" cannot return metallurgy results because those passages are not in that core.
  • The term filter is the right tool for error codes. Semantic search smooths identifiers away; term requires them exactly.

What Cursor gets

All 11 tools, from one config block.

web_search

Searches the agent-first web index and returns metadata-only hits — url, host, snippet, score — never the page body.

get_document

Fetches the full cleaned text of one passage by id — the lazy read path.

similar

More-like-this: given a passage id, returns its semantic neighbours.

browse

Browses without a query: the most recent, or highest-ranked, passages — optionally within one vertical.

why_not

Explains why a URL is or is not in the index: present, removed with a reason, or never admitted.

related

Returns the connected neighbourhood of a passage from the coverage graph — the edges are what to read next.

corroborate

Reports how many distinct hosts independently carry a story — the anti-hallucination check.

authorities

Returns the top passages on a topic ranked by graph centrality — trust-triage a field without reading junk into context.

dossier

Builds a one-hop brief on an entity: mentions, the hosts covering it, and the entities co-mentioned with it.

path

Finds the shortest chain of edges linking two passages — reasoning-path retrieval, or connect-the-dots.

assemble_context

GraphRAG as a service: returns a ready-to-read context pack — corroborated, story-deduplicated, trust-ranked and packed to a token budget, each passage carrying the reason it was included.

Frequently asked questions

Is this better than the built-in web search?

For documentation lookups, usually — you can restrict to docs content type and a specific site, which built-in search does not expose. For general questions the difference is smaller.

Get a key and connect it

10,000 requests a month, no card. The config above works unchanged once you have a key.

Get your API keyMCP server docs