Loading ToolBench
Fast code intelligence engine for AI coding agents — single static binary MCP server
Codebase-memory-mcp has 10 tools with basic naming and descriptions, but significant gaps in schema completeness, parameter documentation, and error handling. Tool names follow verb_noun convention (index_repository, search_graph, get_code_snippet), which is good. However, input schemas visible in the provided code show minimal parameter descriptions. For example, 'index_repository' has 'repo_path' and 'mode' parameters with brief descriptions but no type constraints (enums for mode, path validation rules). Most tools lack documented output schemas, critical for LLM planning. Error handling is not evident in the provided source. The server is written in C with custom MCP implementation, making it difficult to verify full schema compliance without seeing the complete mcp.c registration code. Based on visible evidence, this is a D-grade server with functional but incomplete definitions.
Delete an indexed project
Get high-level architecture information for a project
Retrieve a code snippet from a file
Get the schema of the code graph
Index a repository to build a code graph
List all indexed projects
Query the code graph with structured queries
Full-text search across code files
Search the code graph for symbols, definitions, and references
Output schemas not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., does search_graph return file_path, line_number, symbol_type?). This violates the response-shaper pattern and forces agents to guess at response structure.
Parameter descriptions lack constraints and format guidance. 'mode' in index_repository should specify enum values ('full'|'incremental'). 'query' parameters lack examples of valid syntax. No min/max for line numbers in get_code_snippet.
delete_project is destructive but lacks confirmation/dry-run pattern. No description warns of irreversibility. Agents may accidentally delete indexed projects without safeguards.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | <=2025-11-25 | v2 |
Trace execution paths through the code graph
No error handling guidance visible. Tools do not document what happens on invalid input (e.g., nonexistent project, malformed query syntax, out-of-range line numbers). Agents have no recovery path.
Tool composition risk: search_graph and query_graph appear to overlap in purpose. Descriptions do not clarify when to use each. LLMs will waste reasoning cycles deciding between them.