Loading ToolBench
MCP server for RepoItDown - AST-aware codebase topology for LLM context windows
Single tool 'get_codebase_topology' has a well-structured schema with 4 parameters, all typed and described. Description is 189 chars, within the 10-1024 baseline. However, the tool lacks error handling guidance, output schema documentation, and tool annotations (readOnlyHint). The mode enum is properly constrained, but the tool does not document what fields are returned or how to handle failures. No pagination or result-limiting guidance despite potentially large outputs.
Generates AST-aware codebase topology in Markdown format. Supports multiple processing modes (dump, explore, architect, task) with token-optimized output for LLM context windows.
Output schema not documented. Tool description does not specify what fields are returned (e.g., is it a string, object, array?). LLMs cannot plan downstream operations without knowing the response structure.
No error handling guidance. Description does not explain what happens on invalid repo_path, missing query in task mode, or token budget exhaustion. LLMs have no recovery path.
Missing tool annotations. No readOnlyHint, destructiveHint, or idempotentHint declared. The tool is read-only and idempotent, but this is not signaled to the MCP client.
No pagination or result-limiting strategy documented. Tool accepts max_tokens but does not explain truncation behavior, whether results are sliced or rejected, or how to retrieve remaining data.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
Parameter 'query' is required only in task mode but not marked as conditionally required in schema. LLMs may omit it in other modes without error, or pass it unnecessarily.