Loading ToolBench
Bridges MCP with the AGNTCY multi-agent network (Linux Foundation, 65+ enterprise partners). Supports ACP (Agent Communication Protocol v1.14), SLIM secure messaging, and the AGNTCY Directory for agent discovery. Register agents, discover by capability, create ACP messages, validate protocol compliance, and format SLIM messages for secure real-time agent communication.
Server has 7 tools with basic descriptions and parameter types visible in FastMCP decorators. However, critical gaps undermine quality: (1) descriptions lack actionable context for LLM selection, they read as feature lists rather than 'when to use this' guidance; (2) no output schemas documented anywhere, LLMs cannot plan downstream calls or extract required fields; (3) parameter descriptions are minimal (e.g., 'Agent name' vs 'Agent name (2-50 chars, alphanumeric + hyphens)'); (4) no error handling guidance, tools return dicts but no recovery paths documented; (5) no validation rules stated (e.g., what makes a valid capability string?). Tool names follow verb_noun pattern (good), but lack specificity, 'tool_register_agntcy_agent' vs 'register_agent_in_directory' would clarify scope. Baseline: 549 production tools average 194-char descriptions; these average ~120 chars and lack depth. No evidence of pagination, idempotency, or composition patterns.
Create an ACP-compliant message (Agent Communication Protocol v1.14). ACP is the AGNTCY standard for structured agent-to-agent communication. Message types: request, response, notification, error.
Find AGNTCY agents by capability in the local directory. Searches registered agents by capabilities, name and description. Ideal for task routing and agent selection in multi-agent systems.
Format a SLIM message (Secure Low-latency Interactive Messaging). SLIM is the AGNTCY protocol for secure real-time agent communication. Supports pub-sub, end-to-end encryption and low latency.
Get information about the AGNTCY network and ecosystem. Shows protocol versions, member organizations, SDKs and current stats of the AGNTCY multi-agent network (Linux Foundation project, 65+ members).
List all locally registered AGNTCY agents. Shows all agents in the local AGNTCY directory with optional filter by capability or organization.
No output schemas documented. Tools return `dict` with no field definitions. LLMs cannot plan downstream calls or extract required IDs for chaining.
Parameter descriptions lack actionable constraints. E.g., 'capabilities' is 'List of capabilities', no format, length, or valid examples. 'message_type' lacks enum constraint.
Tool descriptions are feature-focused, not selection-focused. 'Find AGNTCY agents by capability' does not explain WHEN to call this vs list_agntcy_agents, or what the LLM should do with results.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
Register an agent in the local AGNTCY-compatible directory. Creates an AGNTCY Directory Entry with ACP-compatible profile. The agent becomes discoverable via discover_by_capability.
Validate an ACP message against the v1.14 specification. Checks required fields, types and formatting against the AGNTCY Agent Communication Protocol standard.
No error handling guidance. Tools return dicts but no documentation of error cases, recovery paths, or what to do if validation fails.
No pagination support on list/discovery tools. tool_discover_by_capability accepts 'limit' but no offset/cursor. Large result sets will blow context window.