Historical Sverklo performance snapshot
2026-04-08 measurements for Sverklo v0.2.11: search latency, impact-analysis time, cold-start, and DB size on five OSS codebases using one Apple Silicon laptop. These numbers are hardware- and corpus-specific and do not claim current-release performance. For retrieval F1, token economy, and loss cases, see the separately dated May 13 retrieval bench.
Proof caveat: this route records a historical v0.2.11 performance run. It does not prove agent correctness, editing outcomes, or current-release performance. Reproduce against the commit-pinned source and record a new dated result before making a current-release performance claim.
Latency Results
Five real open-source codebases. All measurements on one Apple Silicon laptop with Node 25, Sverklo v0.2.11, and --expose-gc. Results can vary with hardware and corpus. Documented method: node --expose-gc scripts/perf-benchmark.mjs <project paths>.
| Repo | Files | Chunks | Search p50 | Search p95 | Impact | Cold Index | DB Size |
|---|---|---|---|---|---|---|---|
| gin-gonic/gin (Go) | 99 | 1,413 | 11 ms | 12 ms | 0.75 ms | 10.3 s | 4.2 MB |
| microsoft/TypeScript (TS) | 707 | 10,873 | 20 ms | 35 ms | 2.44 ms | 71.1 s | 47 MB |
| nestjs/nest (TS/JS) | 1,709 | 2,976 | 14 ms | 14 ms | 0.88 ms | 22.4 s | 11 MB |
| django/django (Python) | 2,942 | 10,900 | 20 ms | 22 ms | 2.0 ms | 67.9 s | 56 MB |
| facebook/react (TS/JS) | 4,368 | 20,144 | 23 ms | 26 ms | 1.18 ms | 152 s | 67 MB |
MCP Server Comparison
Feature comparison across code intelligence MCP servers. Green = has the feature, dash = does not.
| Server | Hybrid Search | PageRank | Impact Analysis | Diff Review | Cross-Repo | Memory | Local-First | Open Source |
|---|---|---|---|---|---|---|---|---|
| Sverklo | BM25+Vec+PR | Yes | Yes | Yes | GraphQL | Bi-temporal | Yes | MIT |
| Claude Context | BM25+Vec | — | — | — | — | — | Partial | Apache-2.0 |
| Code-Index-MCP | Vec | — | — | — | — | — | Yes | OSS |
| AiDex | Identifier | — | — | — | Link | Session | Yes | MIT |
| Repomix | Pack | — | — | — | — | — | Yes | MIT |
Community submissions welcome — open an issue to add your MCP server to this table.
What We Measure
- Search Relevance — 10 natural-language queries per repo, results manually graded
- Search Latency — p50 and p95 across 9 queries (first dropped as warmup)
- Impact Analysis — time to walk the call graph for the most-referenced symbol in each repo
- Token Efficiency — useful context per token returned (compact mode vs full)
- Cold Index Time — full parse → chunk → embed → PageRank pipeline
Key Findings
- Historical search result: p95 was 12ms on 99 files and 26ms on 4,368 files in this v0.2.11 run.
- Historical impact result: impact analysis was below 3ms across these five measured repositories.
- Historical indexing result: this run observed roughly 7ms per chunk and 152 seconds for the 20,144-chunk React corpus.
- Historical memory result: steady-state RAM was about 200MB; measured peaks during indexing were 400–700MB.
Reproduce It
git clone https://github.com/sverklo/sverklo cd sverklo git checkout fc372029ffdeac4422b4335ecc68ae55e2d38cda npm install && npm run build node --expose-gc scripts/perf-benchmark.mjs <project paths>
Commit-pinned methodology and recorded data: BENCHMARKS.md at fc372029ffdeac4422b4335ecc68ae55e2d38cda