Loading ToolBench
A FastAPI-based server for managing model contexts and sessions with REST API endpoints
This server exposes 11 context/session management tools via FastAPI HTTP. All tools have basic descriptions (10-30 chars) and input schemas with type definitions. However, descriptions are severely underdeveloped, most are single-sentence, generic statements like 'Create a new context' that lack actionable guidance for LLM selection. Parameter descriptions exist but are minimal (e.g., 'Name of the context'). No output schemas are documented. Error handling is present (404 responses) but provides no recovery guidance. Tool naming follows verb_noun convention (create_, get_, update_, delete_, list_) which is correct, but descriptions fail to explain WHEN to use each tool or what distinguishes similar operations. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present despite clear risk levels (WRITE, DESTRUCTIVE, READ_ONLY). The server lacks pagination guidance for list operations, no per-item error reporting for batch-like scenarios, and no documented chaining IDs for multi-step workflows.
Create a new context.
Create a new session.
Delete a context.
End a session.
Get a context by ID.
Get a session by ID.
Get all sessions for a specific context.
List all contexts.
Tool descriptions are generic and under 20 characters. Examples: 'Create a new context', 'Get a context by ID', 'List all contexts'. These lack WHEN to use, prerequisites, or differentiation from similar tools. LLMs cannot reliably select the right tool.
No output schemas documented. Callers cannot know what fields to expect from responses (e.g., does list_contexts return pagination info? What fields are in each context object?). Agents cannot plan downstream tool calls.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
List all sessions.
Update an existing context.
Update an existing session.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. delete_context is marked DESTRUCTIVE but has no annotation. Agents cannot distinguish safe retries from irreversible operations.
list_contexts and list_sessions lack pagination parameters (limit, offset, cursor) and no result count documented. Large result sets will blow context windows. No guidance on result limits.
Error responses (404 'Context not found') provide no recovery guidance. No actionable error messages like 'Try list_contexts() to see available IDs' or suggestions for similar resources.