Loading ToolBench
Personal AI ops agent starter kit — Claude Code edition. Provides Discord slash command integration, scheduled cron job execution, and interactive Claude Code session management with model switching capabilities.
This server exposes 7 tools via Discord slash commands, but lacks proper MCP tool registration. Tool definitions are inferred from Discord handler code rather than explicitly registered with schemas. No input schemas are visible in the source, parameters are parsed from Discord interactions, not declared as JSON Schema. Descriptions exist but are verbose and lack LLM-optimized structure. Critical gaps: no parameter type definitions, no output schemas, no error recovery guidance, and no tool composition support. The codebase is a Discord bot wrapper around Claude CLI, not a proper MCP server.
List all enabled cron jobs
Tail the log for a cron job
Manually trigger a cron job by ID
Send an arbitrary prompt to your agent via claude -p
Show or set the active Claude model. Allows reading the current model or switching between opus, sonnet, and haiku variants.
Post a message to a Discord channel via bot token. Discord max message length is 2000 chars.
Check agent health — uptime, memory, cron status. Run a quick system status check: report memory usage, list any recent errors from memory/errors.md if it exists, and confirm cron runner is healthy.
No JSON Schema input definitions visible. Tools parse Discord interaction options directly; no formal schema registration for MCP.
Tool 'herc' is vague and does not start with a clear action verb. 'herc' is jargon; 'send_prompt_to_agent' or 'run_claude_prompt' would be self-documenting. LLMs cannot infer intent from 'herc'.
No output schemas documented. Tools return free-text or unstructured results (e.g., 'status' returns raw Claude output). LLMs cannot plan downstream calls or extract structured data.
Parameter descriptions lack type and constraint information. 'message' param in 'herc' has no length limit, format, or validation guidance. 'id' param in 'cron run' has no enum of valid job IDs.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 29 | 2026-07-28+ | v2 |
No error recovery guidance. If 'cron run' fails, no actionable error message tells the LLM what to do next. No distinction between retryable and fatal errors.