Loading ToolBench
MCP server that gives AI assistants (Gemini, Claude, Cursor, Windsurf, Copilot) deep understanding of your codebase — 10 tools + persistent memory
CodeAtlas has 10 tools with consistent naming (verb_noun pattern) and reasonable descriptions (avg ~150 chars). All tools have input schemas with Zod validation. However, parameter descriptions are sparse or missing in several tools (e.g., get_insights has empty input {}), and output schemas are not formally documented. Error handling is minimal, most tools return generic 'not found' messages without recovery guidance. The sync_system_memory tool is a WRITE operation but lacks confirmation/dry-run safeguards. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications.
Generate a Mermaid diagram showing how a specific feature flows through the system, including all involved entities and their relationships.
Generate a Mermaid diagram showing the overall system architecture and data flow between major components.
Get import/call/containment/implements relationships between entities. Shows how modules, classes, and functions are connected.
Get all entities (functions, classes, variables) defined in a specific file.
Get AI-generated code insights including refactoring suggestions, security issues, and maintainability analysis.
Get all modules, classes, functions, and variables in the analyzed project. Returns entity type, name, file path, and line number.
get_insights has empty input schema ({}). No parameters documented, making it impossible for LLMs to understand what data it returns or how to invoke it meaningfully.
sync_system_memory is a WRITE operation (modifies persistent files) but lacks confirmation/dry-run safeguards. No tool annotations (destructiveHint) to warn agents of side effects. Agents could accidentally overwrite business rules or changelogs.
Output schemas are not formally documented. Tools return JSON objects but LLMs cannot see the expected field structure, types, or required fields. This forces LLMs to guess at response shape and risks parsing errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
List all projects that have been analyzed by CodeAtlas. Returns project names, paths, and last analysis time.
Search for functions, classes, modules, or variables by name. Supports fuzzy matching.
Sync system understanding to persistent memory files (business-rules.json, change-log.json, conventions.md). Stores key insights and architectural decisions.
Trace the flow of a specific feature through the codebase. Shows all functions, classes, and files involved in implementing a feature.
Error handling is generic. When analysis data is not found, tools return 'No analysis data found. Run CodeAtlas: Analyze Project in VS Code first.' This does not guide LLMs on recovery steps or alternative actions.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications in the spec. LLMs cannot distinguish safe read-only tools from destructive ones without explicit hints.