Bridges MCP with Google's Agent2Agent (A2A) Protocol. Enables agent discovery, registration, task delegation, and inter-agent communication following the A2A standard. Agents can publish Agent Cards, find other agents by capability, and delegate tasks using the A2A task lifecycle.
Server has 6 tools with basic descriptions and parameter types visible in source. However, critical gaps undermine quality: (1) Parameter descriptions are minimal or missing context (e.g., 'agent_card' lacks detail on required fields); (2) Output schemas are not documented, tools return dicts but structure is opaque to LLMs; (3) Error handling is absent, no guidance on recovery, retryability, or actionable error messages; (4) No input validation or constraint documentation (enums, ranges, formats); (5) Tool names are verb-prefixed but descriptions lack WHEN/WHY context needed for LLM selection. Naming is acceptable (tool_* prefix, action verbs), but descriptions average ~100 chars and lack the 50-200 char LLM-optimized range. Per-tool scores range 40-65; average 52.
Create an A2A-compatible Agent Card with skills, capabilities and contact endpoint. An Agent Card is the standard way for agents to advertise their capabilities in the A2A protocol. Other agents can discover and interact with you through it.
Search for agents by capability in the local A2A registry. Finds agents whose skills, description or name match the given capability. Useful for finding the right agent to delegate a task to.
Check the status of a delegated task. Returns the current state, all status transitions and messages for a task.
List all registered agents with their capabilities. Returns an overview of all Agent Cards in the local directory, including names, endpoints and skills.
Register an agent in the local A2A directory. Stores the Agent Card in ~/.a2a-agents/ so other agents can discover it. If an agent with the same name already exists, it will be updated.
No output schemas documented. Tools return dicts but LLMs cannot infer field names, types, or structure. Downstream tool chaining is blind.
Parameter descriptions are sparse or missing context. 'agent_card' param has no detail on required fields (name, url, skills structure). 'input_data' is vague about format/encoding.
No error handling or recovery guidance. Tools return error dicts but do not classify as retryable/user-fixable/fatal or suggest next steps. LLM cannot self-correct.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Create and send a task request following A2A protocol format. Creates a task with taskId, status lifecycle and messages, addressed to a specific target agent.
No input validation or constraint documentation. 'capability' param accepts free-form strings; no enum or pattern guidance. 'endpoint' URL format not validated or documented.
Tool descriptions lack WHEN/WHY context. 'Create an A2A-compatible Agent Card' does not explain when to call this vs register_agent, or what prerequisites exist. LLM selection is ambiguous.