Last reviewed:
Connect your AI coding agent directly to real-time threat intelligence. Search threats, pull detections, query IOCs, and map MITRE techniques without leaving your editor.
The Model Context Protocol (MCP) is an open standard created by Anthropic that defines how AI applications connect to external data sources and tools. Instead of copying data into prompts or writing one-off integrations, MCP provides a single protocol that any AI client can use to discover, read, and invoke tools from any compatible server.
The Threadlinqs MCP Server exposes the entire intelligence platform—threats, detections, IOCs, MITRE mappings, C2 intelligence, attack simulations, and advanced correlations—as 81 callable tools, 25 analyst prompts and a set of browsable resources. Your AI agent gets structured access to the same data analysts use on the web platform, with no API wrangling required.
There are two transports and one catalog. The remote endpoint at https://intel.threadlinqs.com/mcp speaks Streamable HTTP and signs you in with OAuth 2.1, which suits Claude Desktop and Claude.ai custom connectors. The local npm package runs on your machine via npx, talks to your client over stdio, and forwards every request to that same /mcp endpoint with your API key. It opens no ports and needs no webhook configuration.
No global install needed. The package runs via npx on demand. Add it to your AI client configuration and it starts automatically when your agent needs threat intelligence.
{
"mcpServers": {
"threadlinqs-intel": {
"command": "npx",
"args": ["-y", "intelthreadlinqs-mcp"],
"env": {
"THREADLINQS_API_KEY": "tl_your_key_here"
}
}
}
}
# Add the MCP server to Claude Code
claude mcp add threadlinqs-intel \
-e THREADLINQS_API_KEY=tl_your_key_here \
-- npx -y intelthreadlinqs-mcp
{
"mcpServers": {
"threadlinqs-intel": {
"command": "npx",
"args": ["-y", "intelthreadlinqs-mcp"],
"env": {
"THREADLINQS_API_KEY": "tl_your_key_here"
}
}
}
}
Using Claude Desktop or Claude.ai? Add a custom connector with the endpoint https://intel.threadlinqs.com/mcp and complete the OAuth 2.1 sign-in with a Purple or Gold account; there is no key to manage.
For the local package the API key is required — every MCP tool call is authenticated and gated to the Purple and Gold tiers. Generate a tl_ key in your profile settings; the env block above is not optional.
Every tool returns structured JSON. Your AI agent can chain tool calls—search for a threat, pull its detections, look up the associated MITRE techniques, and export a Sigma rule—all in a single conversation turn.
The groups below are a representative selection, not the full inventory. The complete, always-current catalog with input schemas is published at /mcp/catalog.json.
Resources are read-only data your AI agent can address by URI instead of a tool call. Resource templates take an entity id, so a client can open a threat, a CVE or an actor directly.
The server also publishes interactive views under ui://threadlinqs/ (threat feed, MITRE matrix, detections, IOCs, correlation graph and more). Clients that support the MCP Apps UI extension render them inline; other clients ignore them. The full list is in /mcp/catalog.json.
The local package uses the stdio transport—the most widely supported MCP transport layer—and the remote endpoint speaks Streamable HTTP. Any client that implements the Model Context Protocol specification can connect. These are the clients we test against:
Every MCP tool requires authentication. The server is gated to the Purple ($11.99/mo) and Gold tiers — there is no free or Blue-tier MCP access, and no tool answers without a Purple or Gold account, signed in over OAuth 2.1 on the remote endpoint or identified by a tl_ API key on the local package. The public documentation and the tool catalog at /mcp/catalog.json are the only unauthenticated surfaces. MCP calls spend the account’s TL Creds like an API key.
Once the MCP server is connected, your AI agent can answer threat intelligence questions directly:
# Threat research "Search for threats related to Cobalt Strike" "Get full details on TL-2026-0288" "What are the most recent critical threats?" # Detection engineering "Show me all SPL detections for ransomware threats" "Export the Sigma rule for detection DET-0042" # MITRE mapping "What techniques does T1059.001 cover?" "Show MITRE coverage for the Initial Access tactic" # C2 intelligence "List active C2 beacons with Cobalt Strike framework" "Show operator clusters and shared infrastructure" # Vulnerability lookup "Look up CVE-2024-3400 and show linked threats" "What is CWE-79 and how do I mitigate it?"
Package: intelthreadlinqs-mcp on npm Version: 8.2.1 SDK: @modelcontextprotocol/sdk v1.26.0
Transports: remote Streamable HTTP at https://intel.threadlinqs.com/mcp; local stdio proxy (no open ports) Runtime: Node.js 18+ for the local package
Protocol: JSON-RPC 2.0, MCP protocol version 2025-11-25 (2025-06-18, 2025-03-26 and 2024-11-05 still accepted)
Data source: one tool registry in the Threadlinqs platform, backed by Cloudflare D1
Authentication: OAuth 2.1 (PKCE S256) on the remote endpoint; Bearer token via THREADLINQS_API_KEY for the local package
Timeout: 30s per proxied request with automatic abort controller
The local package is stateless and holds no tool logic. It forwards tools/list, tools/call, prompts/* and resources/* to /mcp over HTTPS with your key and returns the response, so both transports expose the same catalog with the same argument names, and the tier gate is enforced in one place. Responses are returned as structured JSON content blocks, ready for the AI model to parse and reason over. Without a network connection, tool listing falls back to the public /mcp/catalog.json and then to a catalog snapshot bundled in the package.
The major releases of the Threadlinqs Intelligence MCP server, oldest first. Tool calls are designed for Purple and Gold accounts only. Before the v8 proxy rewrite, recorded under v8.1.0, the stdio package did not enforce that for every tool.
49 tools on two transports that share one catalog: a remote Streamable-HTTP endpoint at https://intel.threadlinqs.com/mcp with OAuth 2.1 (PKCE S256) for Claude Desktop and Claude.ai, and the local intelthreadlinqs-mcp npm package over stdio, authenticated with a tl_ API key. Both speak JSON-RPC 2.0 and target MCP protocol version 2025-06-18.
The server became open source under the MIT license at github.com/threadlinqs-cmd/intelthreadlinqs-mcp and was listed on the official MCP registry as io.github.threadlinqs-cmd/intelthreadlinqs-mcp. The v7 rebuild made the same 49 tools lean by default: tighter, structured payloads, with richer detail only on request.
The npm package was rewritten as a thin stdio-to-HTTP proxy over /mcp (1,798 lines down to 436), so the worker's registry is the single catalog: both transports expose the same 73 tools with the same argument names. The rewrite closed a gap where roughly twenty stdio tools reached tier-0 REST endpoints without a key. v8.1 added 25 analyst prompts covering all 73 tools, a bundled catalog snapshot so tool listing works offline, and an agent skill at /mcp/skill.md. Read the MCP server v8.1.0 release notes.
The registry grew from 73 to 81 tools: attack-flow, D3FEND and ATLAS tools (get_attack_flow, get_flow_similar_threats, get_flow_coverage, get_d3fend_posture, get_atlas_intelligence), plus get_entity_aliases, get_actor_relationships and list_campaigns. The catalog listed 25 prompts, 14 resources and 3 resource templates. The stdio proxy itself did not change; its bundled catalog snapshot and agent skill were regenerated from the live registry. The current npm release is 8.2.1.
The MCP server is one piece of the full intelligence platform. See how it connects to threat intelligence, detection, and coverage.
Curated threat feed with CVE enrichment, indicators of compromise, and daily debriefs.
31,000+ SPL, KQL, and Sigma detection rules mapped to MITRE ATT&CK.
Every observed technique across 14 tactics, with detection debt scoring.
300+ threat actors profiled with nation-state mapping and correlation.
Atomic red team-style simulations in three flavors for detection validation.
Compare Blue, Red, Purple, and Gold tiers across the full platform.
See how Threadlinqs compares on IOC aggregation and transparent pricing.