Sharing progress on AgentVault, an open-source ecosystem I've been building for AI Agent-to-Agent (A2A) communication. The core client library (agentvault), CLI (agentvault-cli), and server SDK (agentvault-server-sdk) are now available on PyPI! 🚀

The goal is to provide a standardized and secure foundation for AI agents to discover and interact with each other. The current implementation aligns with emerging A2A concepts (JSON-RPC over HTTP/SSE) and includes:

  • agentvault Library: An async Python client with secure credential management (KeyManager supporting env, file, OS keyring) and Agent Card parsing.
  • agentvault-cli: A command-line tool for key config, agent discovery via the registry, and running A2A tasks.
  • agentvault-registry: A FastAPI/PostgreSQL backend serving as the central discovery hub (API + basic UI).
  • agentvault-server-sdk: Helpers (like create_a2a_router) to easily build A2A-compliant agent servers using FastAPI and Pydantic.

By enabling standardized interoperability, AgentVault aims to unlock opportunities for more complex multi-agent workflows, specialized agent services, and a more connected AI ecosystem overall.

The foundation is laid, and the focus now shifts to refinement, more examples, and documentation. Developer feedback is highly welcome as the project and A2A standards evolve!

Explore the initial documentation & components:
https://secureagenttools.github.io/AgentVault/