Loading ToolBench
Agentic repo-to-wiki: scan any repository into a portable SQLite knowledge store with wiki pages, diagrams, and grounded Q&A. Exposes codebase knowledge as MCP tools for AI agents.
Rekipedia provides 14 well-named, read-only tools for codebase analysis with consistent verb-noun naming (ask, get_*, search_*, list_*, detect_). All tools have descriptions (avg ~120 chars, within baseline 34-392 range). Input schemas are present with type definitions and parameter descriptions. However, output schemas are not documented, LLMs cannot predict return structure for planning. Error handling is absent; tools provide no recovery guidance. Parameter descriptions lack format constraints (e.g., depth ranges, query syntax). No tool annotations (readOnlyHint, idempotentHint) despite all being read-only. Composition is strong: tools chain well (e.g., list_wiki_pages → get_wiki_page), and natural identifiers (file paths, symbol names) are accepted. Security is solid for read-only operations but lacks audit logging.
Answer a natural-language question about this codebase, grounded in the scanned wiki pages, symbol index, and source code embeddings. Use this before reading any source file.
Detect uncommitted git changes and return blast-radius impact for each changed file. Returns {changed_files, impacts: [{file, symbols, risk, total_affected, affected_files}]}. Use this before a code review to understand the blast radius of pending changes.
Find which import-graph community a symbol belongs to and list all community members. Uses label-propagation on the relationship graph. Returns {community_id, members: [symbol_names]}.
Get symbols and relationships for a file (partial filename match supported)
List the top N 'god nodes' — symbols with the highest combined in+out degree in the call graph. These are architectural chokepoints that every other module depends on. Returns [{name, in_degree, out_degree, score}].
List architectural chokepoints
Output schemas not documented. LLMs cannot predict return structure (fields, types, nesting) for planning downstream calls or extracting data. E.g., get_impact returns {file, symbols, risk, total_affected, affected_files} but this is not formally declared.
No error handling or recovery guidance. Tools provide no actionable error messages. If a symbol is not found or a file path is invalid, LLMs receive no hint on what to try next (e.g., 'Symbol not found. Try search_nodes() with a partial name.').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2026-07-28+ | v2 |
Blast-radius for a changed file
List untested high-call-count symbols
Get callers and callees for a symbol
BFS transitive impact from a symbol name
Read the full content of a specific wiki page by name or filename. Returns markdown content with symbol cross-references. Use after list_wiki_pages to get detailed documentation for a module or concept.
List all wiki pages generated by rekipedia for this codebase. Returns page names and file paths. Use this to discover what topics are documented before calling get_wiki_page.
Search symbol names (fast indexed lookup)
Find the shortest directed call-path between two symbols in the dependency graph. Useful for understanding 'how does A reach B?'. Returns {path: [A, ..., B], length: N} or {path: null, length: null} if unreachable.
Parameter descriptions lack format constraints. E.g., 'depth' parameters (get_impact, get_transitive_impact) have no stated range (min/max). 'query' in search_nodes lacks syntax guidance. LLMs may pass invalid values without clear bounds.
No tool annotations despite all tools being read-only and idempotent. Missing readOnlyHint and idempotentHint annotations prevent clients from optimizing caching, retry logic, and safety classification.
No pagination support for list_wiki_pages or result-limiting tools. Large codebases may return hundreds of wiki pages or symbols, exhausting context windows. Missing limit/offset parameters and total_count in responses.