jagmarques/asqav-sdk

Python and TypeScript SDKs for verifiable evidence of AI agent actions. Signed receipts, policy enforcement, audit trails. Works with LangChain, CrewAI, MCP.

What it solves

Asqav provides an "evidence layer" for AI agents to solve the lack of accountability and transparency in autonomous systems. It prevents the problem of agents performing dangerous actions without a record, allows for the enforcement of policies before execution, and automates the generation of compliance reports for regulations like the EU AI Act and DORA.

How it works

The SDK allows developers to sign every action an agent takes using ML-DSA-65 (FIPS 204) cryptography. This creates a cryptographically signed audit trail and a verifiable action record.

It supports two data handling modes:

  • Cloud mode: Sends only a SHA-256 hash (fingerprint) of the action context to the Asqav cloud to maintain privacy.
  • Self-hosted mode: Sends the full payload to the server for richer auditing and PII redaction.

Verification can be done via a public URL, an API call, or entirely offline using a standalone verifier to ensure the records are not dependent on Asqav's servers.

Who it’s for

Developers building AI agents that require strict governance, auditability, and regulatory compliance, particularly those operating in highly regulated industries or deploying autonomous coding agents.

Highlights

  • Cryptographic Identity: Assigns agents a verifiable identity and signs actions with ML-DSA-65.
  • Multi-Language Support: Official SDKs for Python and TypeScript.
  • Policy Enforcement: Ability to block dangerous actions before they are executed.
  • Regulatory Readiness: Automatic generation of reports for EU AI Act and DORA.
  • Broad Integrations: Native callbacks for frameworks like LangChain, CrewAI, LlamaIndex, and the Vercel AI SDK.
  • Code Authorship: Specific receipts for coding agents to bind changes to the agent that authored them.

Related