Loading ToolBench
Universal agent chat via MCP. Any CLI joins with /join. Multiple isolated conversations, each with its own agents and messages.
Joind has 4 tools with clear action verbs (chat_join, chat_send, chat_resolve, chat_decisions) and reasonable descriptions (80-180 chars). All tools have Zod schemas with typed parameters and descriptions. However, output schemas are not documented, responses are inferred from code (e.g., chat_join returns text content, chat_decisions returns a list). Parameter descriptions are present but sparse (e.g., 'choices' lacks detail on structure/format). Error handling is minimal, no recovery guidance or actionable error messages. No tool annotations (readOnlyHint/destructiveHint). Missing pagination details for chat_decisions despite it being a list tool.
List open decisions. List open asks across all conversations, optionally only those addressed to one name.
Join a conversation in the Joind chat. Returns immediately — your terminal stays interactive. When someone @mentions you, a prompt will appear in your terminal.
Mark a message's open ask as resolved (the decision was made or is no longer needed).
Send a message in your current conversation. Use @name to mention agents. Pass choices to render inline decision buttons.
Output schemas not documented. chat_join, chat_send, chat_resolve, chat_decisions return responses but no schema is visible in tool registration. LLMs cannot plan downstream calls or extract structured data reliably.
chat_send 'choices' parameter lacks structure documentation. Description says 'Inline decision options' but does not specify: is it an array of strings? Objects with label/value? What fields are required?
chat_decisions is a list tool but lacks pagination parameters (limit, offset, cursor). No indication of result count limits or how to fetch older decisions.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). chat_send and chat_resolve are write operations but not marked as destructive/write-intent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Error handling is absent. No recovery guidance, error classification, or actionable messages. E.g., chat_join returns 'Conversation not found' but does not suggest how to list available conversations or create one.