Capframe

Capability security for AI agents. It maps every tool an agent can reach, mints scoped and revocable capability tokens, and enforces every call against a policy at runtime, with no LLM in the decision path. It sells to engineering and security teams in enterprises and startups that ship AI agents. It is MCP-native, audit-mapped to OWASP, NIST, and MITRE ATLAS, and delivered as open-source Rust binaries and a Python package.

Key features

• MCP-native discovery
• Scoped, revocable capability tokens
• Runtime policy enforcement
• Microsecond capability checks
• Audit-ready reports (OWASP/NIST/ATLAS)
• No LLM in decision path
• Findings.v1 JSON schema
• CLI and pip install