Support A2A agents

[mehdihadeli]

Hi,
Do you have a plan for supporting A2A agents as well?

[chopmob-cloud]

AlgoVoi runs A2A in production alongside MCP -- happy to share the routing shape.

AlgoVoi's payment gateway supports both MCP (29 tools) and A2A protocol simultaneously. The routing distinction in practice: MCP handles tool invocation from a single orchestrator, A2A handles agent-to-agent payment delegation where the payer agent and the payee agent may be operated by different parties and need to exchange AgentCards, verify payment capability, and settle on-chain.

For mcp-gateway specifically, the A2A support question comes down to two surfaces:

1. AgentCard routing -- A2A agents advertise capability via /.well-known/agent.json. A gateway that can proxy this endpoint alongside MCP server/discover gives an orchestrator a unified discovery surface.
2. Task lifecycle vs tool call -- A2A uses a task state machine (submitted / working / completed / failed) rather than a synchronous tool response. The gateway needs to handle async task polling and streaming TaskUpdate events alongside synchronous MCP calls.

AlgoVoi's live A2A AgentCard is at https://api.algovoi.co.uk/.well-known/agent.json -- it exposes payment capability fields, supported protocols (x402, A2A, AP2, MPP), and the ZKP trust credential endpoint. Happy to sketch the gateway routing configuration if it is useful for the A2A support design.

AlgoVoi (chopmob-cloud) -- docs.algovoi.co.uk/a2a-integration

[shadowhunter-92]

Ran into the same need. I've been prototyping a small mesh that translates MCP <-> A2A (and ACP / OpenAI & Gemini function-calling / AGNTCY) through one canonical hub-and-spoke layer — every adapter validated against the protocol's real official SDK — plus identity (Ed25519 DIDs) / per-agent budgets / a hash-chained tamper-evident audit log on top.

Code + a one-screen demo here: https://github.com/shadowhunter-92/agentbridge (run python examples/demo_story.py to see the same live MCP tool reached from 6 protocols, with governance enforcing identity / budget / audit in the call path).

Genuinely curious what your A2A use-case looks like — are you bridging out to specific A2A agents, and do you also need governance (cost control / audit trail) around those calls, or just the translation? Asking because I keep hearing those two needs together but want to validate.