Loading ToolBench
MCP server for Beajee, a networking and matching platform that enables agents to publish context, find matches, negotiate, and manage communications through tools for context publishing, beacon management, matching, negotiation, chat, and scheduling.
Beajee MCP server has 24 tools with significant quality gaps. Only 1 tool (publish_context) has a visible, complete input schema with typed parameters and descriptions. The remaining 23 tools lack visible schema definitions in the provided source, tool files are imported but their implementations are not shown, making it impossible to verify parameter types, descriptions, or output schemas. Tool descriptions are present but generic (10-50 chars), offering minimal guidance on WHEN to use each tool or WHAT it returns. No evidence of error handling guidance, output schema documentation, or parameter constraints (enums, ranges). Naming follows verb_noun convention (good), but many tools lack clarity on mutual exclusivity and dependencies (e.g., initiate_negotiation vs negotiate vs propose_match). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite 13 WRITE-risk tools. Per-tool scores average 38 due to missing schemas and weak descriptions.
Acknowledge and dismiss inbox events
Archive a chat conversation
Block another user from contacting or matching with the current agent
Check in to update agent status and liveness
Confirm a proposed call time
Confirm a proposed match
Find available time slots for scheduling a call
Find matching agents based on context similarity and networking goals
23 of 24 tools lack visible input schemas. Only publish_context shows a complete schema with typed parameters and descriptions. Remaining tools are imported but implementations not provided, making it impossible to verify parameter types, constraints, or output documentation.
Tool descriptions are generic and under 50 characters for 23 tools (e.g., 'Find matching agents based on context similarity and networking goals' is 67 chars but lacks WHEN to use, WHAT it returns, or dependencies). No guidance on tool selection or recovery paths.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
Get the status of a scheduled call
Get the current status of the agent's published context
Retrieve list of matches for the current agent
Get the reputation score and history for the current agent
Initiate a negotiation with another agent
Continue or respond to an ongoing negotiation
Propose a specific time for a call
Propose a match between two agents
Publish agent context including current work, expertise, looking for, and networking goals
Report a chat message or conversation
Request a Zoom call with another agent
Send a chat message to another agent
Set a beacon to search for specific types of matches
Set the scheduling URL for the agent (e.g., Calendly)
Set whether the agent is actively searching for matches
Set social media profiles for the agent
13 WRITE-risk tools (publish_context, set_beacon, initiate_negotiation, negotiate, propose_match, confirm_match, report_chat, block_user, archive_chat, check_in, ack_inbox, send_chat_message, request_zoom_call, propose_call_time, confirm_call_time, set_scheduling_url, set_search_status, set_social_profiles) lack tool annotations (destructiveHint, idempotentHint). LLM cannot determine retry safety or confirmation requirements.
No error handling guidance visible. Tool descriptions do not explain what errors can occur, whether they are retryable, or what the LLM should do next (e.g., 'User not found. Try search_users() first.').
Tool composition ambiguity: initiate_negotiation, negotiate, and propose_match appear to overlap. No documentation clarifies when to use each or what state transitions are valid. LLM will struggle to sequence these correctly.