Skip to main content
Connect your AI assistant to Sixtyfour search, research, and docs through two hosted MCP servers:

Setup

The sixtyfour-intelligence server accepts OAuth sign-in with your Sixtyfour account or an API key from Get an API Key. Send the key in an x-api-key header. A key in the URL (?api_key=) is refused, because URLs end up in logs and shared config files; a client that can’t set headers signs in with OAuth instead. Usage is billed to the organization of the signed-in user or key.

Cursor

Install in Cursor Or add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (project) manually. Replace YOUR_API_KEY with your key.
To sign in with OAuth instead, omit headers. Cursor opens a browser sign-in on first use:
Cursor limits tool names to 60 characters and shortens longer names with a hash suffix. If Cursor warns that a tool name is too long, the tool still works.

VS Code

Install in VS Code Or add to .vscode/mcp.json in your project root manually. Replace YOUR_API_KEY with your key.

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json. Replace YOUR_API_KEY with your key.
After saving the configuration, reload your MCP servers or restart your IDE.

Claude Code

Run these commands in your terminal. Add --scope user to make a server available across all projects. Docs MCP
Intelligence MCP: OAuth (recommended) Claude Code opens a browser sign-in on first use:
Intelligence MCP: API key

Claude Desktop

In Claude Desktop, add two custom connectors: Claude Desktop handles authentication for the intelligence server via OAuth on first connection.

Codex

Install via the Sixtyfour Codex plugin, or add the servers to ~/.codex/config.toml manually. Plugin install Register the marketplace, then install the plugin. Start a new Codex session after installing: plugins load at session start. Codex CLI:
Check installed plugins with codex plugin list. To install this plugin in the ChatGPT desktop app instead of the CLI, see ChatGPT.
Which ChatGPT plans and apps can install plugins is set by OpenAI; see Plugins in ChatGPT. On Enterprise, Edu, and Business workspaces, an admin must allow installing marketplace plugins under Workspace Settings → Plugins before it appears for members.
On first use, authenticate the intelligence server. Run codex mcp login sixtyfour-intelligence when prompted.
Manual config Add the servers to ~/.codex/config.toml, then restart Codex and sign in when prompted. Docs MCP:
Intelligence MCP:
Reset a stuck sign-in session The manual config.toml + mcp-remote method caches its OAuth session in ~/.mcp-auth, separate from Codex’s own session state. If sign-in stops working, clear the cached session for the intelligence server:
Restart Codex. It prompts you to sign in again via browser. This only affects the Sixtyfour session; other mcp-remote-based connections are untouched.
To avoid clearing the cache altogether, switch to the plugin install method, which supports codex mcp login/codex mcp logout for reconnecting.

ChatGPT

The Sixtyfour plugin is published in the ChatGPT Plugins Directory, the same plugin installed via the CLI marketplace command in Codex. View in ChatGPT Plugins
  1. Open the listing above, or turn on Work mode in ChatGPT (web or desktop app) and search “Sixtyfour” in Plugins.
  2. Select the plugin and install it.
  3. Authenticate the intelligence server when prompted.
See Codex for plan requirements and workspace admin settings that gate plugin installs.

Documentation MCP reference

The sixtyfour-docs server gives your AI assistant direct access to Sixtyfour API references, guides, and code examples. Example prompts
  • “How do I use the company-intelligence endpoint?”
  • “What parameters does find-email accept?”
  • “Show me an example of using webhooks with Sixtyfour.”

Intelligence MCP reference

The MCP tools keep their own names. A _start tool returns a task_id, not a job_id, and its status tool reports completed, failed, cancelled, terminated or timed_out. For the REST names, see Async Jobs. The sixtyfour-intelligence server exposes the tools below, grouped by category. See Credit Costs for what each tool costs. Each search response returns charge_credits.
Tools ending in _start create billable work. A start call that times out may still have created the job, so do not repeat it automatically. Retry sixtyfour_atlas_start only with the same request_id and inputs.
Account Company search People search Result handles and cursors expire after 20 minutes; rerun the search to get new ones. Set count_only to true on sixtyfour_people_search to return total_available without profiles, result handles, or pagination data. When an exact count is unavailable, total_available is null:
Similar people (lookalike) Pass seeds as linkedin_urls (1–1000 LinkedIn profile URLs) or csv_text (the full text of a CSV with a LinkedIn URL column, up to 1000 rows), not both. The seeds define the audience; use guidance only for hard requirements, such as "only people in the US". Seed profiles never appear in the results. A search typically takes 1–3 minutes. Poll with sixtyfour_find_similar_status, not sixtyfour_enrichment_job_status; set wait_seconds (up to 25) to wait server-side between polls. Page the results by search_id, then by the returned next_cursor. Async research and contact lookup The completed response includes result, charge_credits, and task_type. The intelligence start tools accept the micro, low, medium, and high tiers; for xhigh research, use the Atlas tools below. Bulk intelligence Pass the entities as rows (a list of identifier objects, such as name, company, and LinkedIn URL) or csv_text (CSV text with a header row), not both. struct is required: each key becomes an output field, described in plain English or as {"description": "...", "type": "str"}. The start response includes estimated_cost_credits, a lower bound on the job’s cost. See Credit Costs. Poll with sixtyfour_enrichment_job_status. The completed response includes signed download URLs for the result files under results, and charge_credits. Rows pass through the model’s context, so use these tools for up to a few hundred rows. For larger files, use the web app or the Bulk Intelligence API. Age verification Poll with sixtyfour_verify_age_status, not sixtyfour_enrichment_job_status. The completed response includes date_of_birth, age, evidence, and sources; see Age Verification. Atlas investigations These tools appear only when your organization has Atlas (xhigh People Intelligence) enabled. See Atlas Integrations for how they fit together. Investigations also appear under Atlas in the web app. Feedback See Send Feedback for worked examples. Webhook testing Both tools are free and run no research, monitor or workflow. sixtyfour_send_test_webhook counts toward the POST /webhooks/test-events rate limit. Each call sends a new event, so read delivered in its response instead of retrying. IDs in a sample event do not belong to real jobs or monitors. Example prompts
  • “Search for US SaaS companies with 100–5000 employees.”
  • “Research Sixtyfour: return headquarters, headcount, funding stage, and key executives.”
  • “Research this person: Saarth Shah, CEO & Co-Founder at Sixtyfour.”
  • “Find the CTO and VP of Engineering at Sixtyfour.”
  • “Find the work email for Saarth Shah at Sixtyfour.”
  • “Find a phone number for Jane Doe, VP of Sales at Acme Corp.”
  • “Who owns the email jane@acme.com?”
  • “Verify the age of Jane Doe, who works at Acme Corp.”
  • “Find 50 more people like the ones in this CSV.”
  • “Find the work email and GitHub profile for each of these 20 people.”
  • “Open my latest Atlas investigation and summarize the findings.”
  • “Send a test monitor notification to my webhook receiver.”
  • “How many Sixtyfour credits do I have left?”