generated: '2026-09-06'
method: probed
source: >-
  https://github.com/diffbot/diffbot-mcp (README, pyproject.toml) and a live anonymous
  MCP session against https://mcp.diffbot.com/mcp — initialize + tools/list, 2026-09-06.
provider: Diffbot
providerId: diffbot
name: Diffbot MCP Server
description: >-
  Diffbot's first-party MCP server. Exposes the Extract, Web Search, Knowledge Graph
  (Enhance / DQL), Crawl and Natural Language surfaces as seven agent tools. Diffbot
  operates a hosted remote endpoint and also ships the same server as a local stdio
  process from the diffbot-mcp repository — the repo deploys directly to the remote.
status: published
deployment:
  mode: both
  endpoint: https://mcp.diffbot.com/mcp
  install: python3 diffbot_mcp_server.py
  package: https://github.com/diffbot/diffbot-mcp
  auth: api-key
  verified: probed
  note: >-
    Remote: an MCP client POSTs to https://mcp.diffbot.com/mcp with the Diffbot token
    supplied as a ?token= query parameter. Local: clone the repo, `pip install .`, then
    run diffbot_mcp_server.py over stdio with DIFFBOT_TOKEN in the environment; a Dockerfile
    is provided. Both are documented by Diffbot in the repository README. The package is
    NOT on PyPI (404 as of 2026-09-06), so `pip install diffbot-mcp` does not work.
x-evidence:
  fetched: '2026-09-06'
  initialize:
    url: https://mcp.diffbot.com/mcp
    method: POST
    http_status: 200
    content_type: text/event-stream
    server_info:
      name: Diffbot MCP Server
      version: 2.14.7
    protocol_version: '2025-06-18'
  tools_list:
    url: https://mcp.diffbot.com/mcp
    method: POST
    http_status: 200
    tool_count: 7
    file: mcp/diffbot-mcp-tools.json
    note: >-
      tools/list answered anonymously after the standard initialize handshake, so the real
      inputSchema for every tool was captured without a token. Tool CALLS still require a
      Diffbot token. prompts/list and resources/list both returned empty arrays.
  well_known:
    probes:
      - url: https://mcp.diffbot.com/.well-known/oauth-protected-resource
        http_status: 404
      - url: https://mcp.diffbot.com/.well-known/oauth-authorization-server
        http_status: 404
    note: >-
      No OAuth discovery documents — the server authenticates with a Diffbot API token,
      not an OAuth flow, so RFC 9728 metadata is correctly absent rather than missing.
transport:
  - streamable-http
  - stdio
runtime: fastmcp 2.14.7
authentication:
  type: api-key
  parameter: token
  location: query (remote) / DIFFBOT_TOKEN environment variable (local stdio)
  signup: https://app.diffbot.com/get-started
tools:
  - name: extract
    summary: >-
      Fetch a URL and return structured JSON or markdown. Positioned by Diffbot as a
      replacement for a generic web_fetch tool, optimized for token economy.
    required: [url]
    parameters: [url, page_type, format]
    output_schema: true
  - name: search_web
    summary: >-
      Web search over Diffbot's own index, reranked for factual relevance and primary
      sources rather than popularity. Returns markdown content chunks for citation.
    required: [query]
    parameters: [query]
    output_schema: true
  - name: enhance
    summary: >-
      Look up an organization or person in the Knowledge Graph by name, URL, location,
      email, employer, title or school and return the full entity record.
    required: [type]
    parameters: [type, name, url, location, email, employer, job_title, school]
    output_schema: true
  - name: resolve_entities
    summary: >-
      Identify named entities in a block of text and resolve each to a Knowledge Graph
      entity with confidence, salience and sentiment.
    required: [text]
    parameters: [text, lang]
    output_schema: true
  - name: crawl
    summary: >-
      Crawl a site and extract every page it visits. Runs as a background job; the tool
      also manages and retrieves those jobs.
    required: []
    parameters: [action, url, job_name, hops, max_to_crawl, max_to_process, restrict_domain, url_crawl_pattern, url_process_pattern, obey_robots, use_proxies, crawl_delay, page_type, format, max_results, offset]
    output_schema: true
  - name: dql
    summary: >-
      Query the Knowledge Graph with DQL for set-shaped questions over organizations,
      people, articles and products.
    required: [query]
    parameters: [query, size, offset, fields, hits_only]
    output_schema: true
  - name: dql_ontology
    summary: >-
      Look up entity types, fields, taxonomies and enums so a dql query can be written
      against field paths that actually exist.
    required: [action]
    parameters: [action, name, search, limit]
    output_schema: true
documentation:
  - https://github.com/diffbot/diffbot-mcp
  - https://www.diffbot.com/docs/interfaces/agent-skills
maintainer:
  note: >-
    The repository README names @jeromechoo as the builder and maintainer, with
    sales@diffbot.com as the faster support path.