Getting Started

Precursor Intelligence MCP Server

The Precursor Intelligence MCP server gives any MCP-compatible client (Claude Desktop, Claude Code, Cursor, Zed, or your own agent) direct, typed access to the same threat-intelligence dataset that powers the REST API.

It is a hosted, remote MCP server. There is nothing to install. Point your client at the endpoint, add your API key, and start calling tools.

POST
https://mcp.precursorintelligence.com

What you can do with it

The server exposes 41 tools spanning the full Precursor Intelligence dataset:

  • CVE detail, EPSS scoring, CISA KEV status, and notable events
  • CPE vendor, product, and version lookups with full-text product search
  • CWE catalogue
  • MITRE ATT&CK tactics and techniques
  • Atomic Red Team test library
  • Threat actor profiles, including industry-targeting lookups
  • IOC intelligence (1M+ records) with substring and exact-match lookup
  • Malware sample metadata
  • SSL and JA3 blacklists
  • Shadowserver honeypot telemetry (per-CVE, per-vendor, and daily aggregates)
  • AI-generated remediation summaries
  • Account, credit balance, and data-freshness introspection

A complete list grouped by domain lives on the Tools page.

How it works

The server speaks JSON-RPC 2.0 over HTTP POST, implementing the MCP Streamable HTTP transport at protocol version 2025-06-18. Every call:

  1. Validates your bearer token.
  2. Charges 1 credit on tools/call (account_status is free).
  3. Runs the query against the dataset.
  4. Returns a standard MCP CallToolResult whose single text block is a JSON-serialised payload, with a _meta.precursor block carrying the request ID, credit balance, and tool metadata.

If the underlying query fails or times out, the credit is refunded automatically.

Tools-only server

The server advertises only the tools capability. It does not expose resources, prompts, sampling, or server-initiated notifications. Clients that auto-discover capabilities will see tools.listChanged: false.

Where to go next