What it does
The server exposes the MITRE ATLAS (Adversarial Threat Landscape for AI Systems) catalogue through MCP tools. The publisher describes it as a set of tactics and techniques for attacking AI systems. The documentation names three tools: query_atlas_technique, list_atlas_tactics and case_study_lookup.
A developer or security reviewer might use it to have an assistant look up adversarial AI techniques, browse tactics, or retrieve case studies while assessing the risks of an AI system.
It is installed as a Python package and run over stdio transport. It is added to an MCP client configuration, such as the claude_desktop_config.json file used by Claude Desktop. Once configured, the assistant calls the tools when asked. The free tier is MIT-licensed and needs no API key.
Limitations and requirements
The free tier is limited to 10 calls per day. Unlimited use and signed attestations are offered only under paid subscriptions. Running it requires an MCP-compatible client, such as Claude Desktop, and a Python environment for installation.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.CSOAI-ORG/mitre-atlas-mcp
MITRE ATLAS — Adversarial Threat Landscape for AI Systems. Tactics + techniques for attacking AI...
v1.0.9As shown on registry.modelcontextprotocol.io when we last read it