/* comparison */

Sverklo vs codebase-memory-mcp

Both are code intelligence MCP servers that index your codebase locally for AI agents. Sverklo combines hybrid search with PR review workflows and persistent memory. codebase-memory-mcp is a single C binary that builds a knowledge graph with sub-millisecond query speed across 66 languages. This page is an honest comparison to help you pick the right tool for your project.

Feature comparison

Sverklo codebase-memory-mcp
License MIT Free for all use MIT Free for all use
GitHub stars Growing 1.6k
Install npm install (Node.js, pure JS, no native bindings) Single C binary (no runtime dependencies)
Search Hybrid BM25 + vector embeddings + PageRank scoring Knowledge graph queries, sub-ms response time
Memory system Git-pinned (memories tied to SHA, staleness detection) Knowledge graph acts as persistent memory
PR review Diff-aware with risk scoring and CI gate (--fail-on) Not included
Blast-radius analysis Yes (trace impact of changes through dependency graph) Not included
Languages 10 via tree-sitter 66 via tree-sitter (C parser)
Indexing speed Incremental reindex with live file watching Linux kernel in ~3 minutes
Web dashboard Yes No
GitHub Action Yes (open-source) No
Runtime Node.js None (static C binary)

Public benchmark coverage

Sverklo is on the public 5-baseline retrieval benchmark — 180 hand-verified tasks across 6 OSS codebases (express, lodash, sverklo, requests, flask, fastapi). Sverklo F1=0.58 leading overall; P1 is 0.63, P4 is 0.84, and jcodemunch uniquely reaches P5 recall 1.00 with F1 0.34. codebase-memory-mcp is not on the public bench yet — we'd add it cheerfully if a maintainer opens a PR adding the baseline. The methodology repo at github.com/sverklo/sverklo-bench documents the contract; the auto-bench CI runs against the express slice and posts a results-table comment within ~10 minutes.

In the meantime, the comparison above is feature-shaped, not bench-shaped — we're describing what each tool exposes, not how each tool scores on a common eval. Treat the above as positioning, not measurement.

When to use each

Choose Sverklo when

  • Your workflow includes PR review with automated risk scoring
  • You need hybrid search that combines text, vector, and graph signals
  • Git-pinned memory with staleness detection matters for your team
  • Blast-radius analysis is important for understanding change impact
  • You want a web dashboard for browsing code intelligence
  • You need a CI-integrated GitHub Action for automated review

Choose codebase-memory-mcp when

  • You work with many languages (66 supported vs 10)
  • You want a single binary with zero runtime dependencies
  • Raw indexing speed is critical (massive codebases)
  • Sub-millisecond query response time matters
  • You prefer minimal tooling: one binary, one knowledge graph
  • You want to avoid Node.js as a dependency

Architecture notes

Sverklo uses SQLite with a hybrid retrieval pipeline (BM25 + ONNX embeddings + PageRank graph centrality). It includes diff-aware PR review, git-pinned memory, and a web dashboard. The trade-off: it requires Node.js, and the parser tiers are 10 first-class + 14 regex-fallback (24 total) versus codebase-memory-mcp's 66 first-class.

codebase-memory-mcp uses a single C binary that builds a knowledge graph from your codebase using tree-sitter parsing across 66 languages. Its focus is speed and simplicity: no runtime dependencies, sub-millisecond queries, and the ability to index massive codebases like the Linux kernel in minutes. The trade-off: no PR review, no git-aware memory, and no web interface.

Try Sverklo

MIT licensed. One command. Works with every MCP client.

click to copy