VeriSwarm is the trust layer for AI-agent ecosystems. We ingest signed behavioral events, compute deterministic trust scores, and return real-time policy decisions your platform can enforce. No heuristics. No guesswork. Just math.

Score — Every agent gets four trust dimensions: identity confidence, risk, reliability, and autonomy. Scores are deterministic, explainable, and built from cryptographically signed behavioral evidence.

Decide — Call one endpoint before any sensitive action. Get an instant allow, review, or deny verdict based on the agent's trust tier and your platform's policy profile.

Govern — Three product pillars extend the core scoring engine:

• Guard — Security scanning, credential leak detection, and an emergency kill switch for compromised agents
• Passport — Verified agent identity, signed capability manifests, and scoped delegation grants
• Evidence — Immutable hash-chained audit ledger with cryptographic integrity verification

MCP Server — Any MCP-compatible agent connects with 3 lines of config. 25 tools for scoring, reporting, Guard, Passport, and Evidence.

Framework Adapters — Drop-in auto-instrumentation for LangChain, with CrewAI and Claude Agent SDK coming soon.

Portable Credentials — Agents carry signed JWT attestations proving their trust status to any platform. Verify offline via our public JWKS endpoint.

Embeddable Trust Badges — Show real-time trust scores on any page with a single <img> tag.

Our SDKs and MCP server are open source under the MIT license. The platform is hosted at veriswarm.ai.

Event ingestion is always free and unlimited.