Problem or Use Case
Is your feature request related to a problem? Please describe.
Currently, Hermes Agent operates as a standalone agent without native interoperability with other AI agent systems. While Hermes excels at individual task execution, it cannot easily delegate tasks to or collaborate with external agents that may have specialized capabilities (e.g., a dedicated research agent, code reviewer, or data analysis agent).
Describe the solution you'd like
Implement support for Google's Agent-to-Agent (A2A) Protocol — an open standard for agent interoperability. This would allow Hermes to:
- Discover other A2A-compatible agents via
/.well-known/agent-card.json
- Delegate tasks to specialized agents using JSON-RPC 2.0 methods like
tasks/send, tasks/get, tasks/cancel
- Receive tasks from external agents, making Hermes accessible as a service
- Stream updates via Server-Sent Events for long-running collaborative workflows
Key A2A methods to support:
tasks/send & tasks/sendSubscribe — submit tasks
tasks/get — check task status
tasks/cancel — abort tasks
tasks/resubscribe — reconnect to task streams
- Agent Card discovery at
/.well-known/agent-card.json
Describe alternatives you've considered
- Custom webhook integrations (fragile, non-standard)
- MCP (Model Context Protocol) — great for tools, but not designed for agent-to-agent communication
- Direct API calls — requires tight coupling and prior knowledge of endpoints
Additional context
- Reference: https://github.com/google/A2A
- Competitor advantage: NullClaw already implements A2A v0.3.0, giving it a head start in multi-agent orchestration scenarios
- Use case example: A user asks Hermes to "Research X and create a presentation." Hermes could delegate research to an A2A-enabled research agent, then compile results into a presentation — all transparently.
This would position Hermes as not just a powerful single agent, but as a central orchestrator in a multi-agent ecosystem.
Thank you for considering this! 🙏
Proposed Solution
I propose integrating the A2A protocol into the Hermes Agent core as a native communication layer. Here is how it should be implemented:
Configuration:
Add an a2a section to the config.yaml (or equivalent config file) to enable/disable the A2A server and define the agent's identity.
Example:
CLI Interaction:
Introduce a new command or flag to interact with external agents:
hermes chat --delegate "https://research-agent.com"
Or a specialized tool that the agent can call internally:
call_remote_agent(endpoint, task)
Discovery Mechanism:
Hermes should automatically look for /.well-known/agent-card.json when a URL is provided as a collaborator.
Protocol Support:
Implement a JSON-RPC 2.0 handler for the tasks/ namespace.
Use SSE (Server-Sent Events) for the tasks/sendSubscribe method to ensure real-time updates for long-running tasks.
Alternatives Considered
No response
Feature Type
New tool
Scope
None
Contribution
Problem or Use Case
Is your feature request related to a problem? Please describe.
Currently, Hermes Agent operates as a standalone agent without native interoperability with other AI agent systems. While Hermes excels at individual task execution, it cannot easily delegate tasks to or collaborate with external agents that may have specialized capabilities (e.g., a dedicated research agent, code reviewer, or data analysis agent).
Describe the solution you'd like
Implement support for Google's Agent-to-Agent (A2A) Protocol — an open standard for agent interoperability. This would allow Hermes to:
/.well-known/agent-card.jsontasks/send,tasks/get,tasks/cancelKey A2A methods to support:
tasks/send&tasks/sendSubscribe— submit taskstasks/get— check task statustasks/cancel— abort taskstasks/resubscribe— reconnect to task streams/.well-known/agent-card.jsonDescribe alternatives you've considered
Additional context
This would position Hermes as not just a powerful single agent, but as a central orchestrator in a multi-agent ecosystem.
Thank you for considering this! 🙏
Proposed Solution
I propose integrating the A2A protocol into the Hermes Agent core as a native communication layer. Here is how it should be implemented:
Configuration:
Add an a2a section to the config.yaml (or equivalent config file) to enable/disable the A2A server and define the agent's identity.
Example:
CLI Interaction:
Introduce a new command or flag to interact with external agents:
hermes chat --delegate "https://research-agent.com"
Or a specialized tool that the agent can call internally:
call_remote_agent(endpoint, task)
Discovery Mechanism:
Hermes should automatically look for /.well-known/agent-card.json when a URL is provided as a collaborator.
Protocol Support:
Implement a JSON-RPC 2.0 handler for the tasks/ namespace.
Use SSE (Server-Sent Events) for the tasks/sendSubscribe method to ensure real-time updates for long-running tasks.
Alternatives Considered
No response
Feature Type
New tool
Scope
None
Contribution