What did Hlido score Srclight?

Srclight scored 72/100 (STEADY) in Hlido's independent, hands-on review.

How much does Srclight cost?

Hlido recorded the pricing model as: Open source. Derived from evidence Hlido holds, not vendor-supplied; verify current prices on the vendor's pricing page.

Does any vendor pay Hlido for placement?

No. Hlido takes no money from the agents it rates — scoring weights stay private and the evidence behind every verdict is public.

Srclight

Name: Srclight review
Item: Srclight
Rating: 72
Author: Hlido Editor

Coding · tested 2026-09-01 · by the Hlido desk, not the vendor

In short: Local-first code-indexing MCP server for AI agents — hybrid FTS5 + tree-sitter + embeddings in a single SQLite file, fully offline, 25 tools.

5 PASS · 0 FAIL of 5 public-surface claims

Visit Srclight → scorecard.json

Quick answer

Srclight scores 72/100 (STEADY) on Hlido’s independent, hands-on test (reviewed 2026-09-01). STEADY (72), medium confidence. Pricing: Open source (free entry point documented).

Hlido Editor · 2026-09-01

Srclight targets a real and growing pain: coding agents burn context and latency making dozens of grep/glob calls to understand a repo, and Srclight offers to replace that with a precise, structured index they can query over MCP. The engineering choices on the public surface are tasteful and coherent — a single SQLite file per repo (no external database, no Docker, no cloud), tree-sitter for precise symbol extraction plus call graphs, and a hybrid search that fuses FTS5 keyword matching with semantic embeddings via reciprocal rank fusion. The privacy posture is a genuine differentiator: embeddings run locally through Ollama, so code never leaves the machine, and it explicitly supports air-gapped and private repos. Twenty-five MCP tools, a one-line pip install, multi-repo workspaces and an optional GPU-accelerated vector cache round out a surprisingly complete story for a young project. The caveats are the caveats of youth and of a public-surface read: the star count is small (early adoption), the performance claims (sub-3ms semantic queries, 60-second setup) are documented rather than independently benchmarked here, and the local-embeddings path adds an Ollama dependency and hardware expectations that the landing page soft-pedals. For agent builders who want retrieval that stays on their own hardware, it is a well-conceived, honestly-scoped tool worth watching.

Why STEADY

STEADY (72), medium confidence. It clears the STEADY bar on a clear, well-argued value proposition (structured index over MCP to cut agent grep/glob overhead), a privacy-first local/offline architecture that is a real differentiator, and a genuinely complete feature surface for its age (25 tools, hybrid search, multi-repo, MIT, PyPI). Kept out of the higher band because adoption is still early (small star count) and the notable performance claims are documented but not independently verified in this public-surface review.

Public-surface checklist

PASS Homepage loads (required)
PASS Primary value prop (required) — 'Deep code indexing for AI agents'
PASS Cta present (required) — 'pip install srclight' / GitHub / PyPI
PASS Pricing or access — Open-source MIT on PyPI; free
PASS Evidence or demo — 3-step get-started (install/index/serve) with commands shown

What we saw

2 screenshots captured by the Hlido engine during the reviewed run (run-3c956c4c026024f5-srclight-dev). Our own captures — not vendor marketing material.

Srclight — run screenshot 1 (home.png) — home.png

Srclight — run screenshot 2 (page_.png) — page_.png

What it does well

Replaces dozens of agent grep/glob calls with a precise, structured, queryable index over MCP
Fully local and offline: local embeddings via Ollama, so code never leaves the machine (air-gapped/private-repo friendly)
Minimal footprint — a single SQLite file per repo, no external DB, Docker or cloud services
Hybrid retrieval done right: FTS5 keyword + semantic embeddings fused via reciprocal rank fusion, plus tree-sitter symbol/call-graph extraction
Complete surface for a young project: 25 MCP tools, one-line pip install, multi-repo workspaces, MIT license on PyPI

What it fails at

Early adoption — small GitHub star count and limited independent track record
Headline performance claims (sub-3ms semantic queries, 60-second setup) are documented, not independently benchmarked here
Local-embeddings path adds an Ollama dependency and real hardware expectations the landing page underplays
Public surface is a single landing page; depth of the 25 tools is described more than demonstrated

Best for

Agent builders who want code retrieval that stays entirely on their own hardware
Private-repo, air-gapped or compliance-sensitive teams that can't send code to a cloud indexer
Claude Code / Cursor / Windsurf users wanting to cut token spend on repository exploration

Not recommended for

Teams that prefer a fully managed, hosted code-intelligence service with no local setup
Users unwilling to run Ollama / provision local embedding compute
Anyone needing a proven, large-scale production track record today

Pricing & access

ModelOpen source
Free entry pointYes — a free tier or open-source edition is documented
Pricing findable on the public surfacePASS Open-source MIT on PyPI; free (tested 2026-09-01)

Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-09-01.

Compared to

Sourcegraph Cody local-offline-mcp-retrieval
Sourcegraph Cody is a mature, largely hosted code-intelligence platform; Srclight is a young, local-first indexing layer exposed purely over MCP. Choose Cody for enterprise-scale hosted code search; choose Srclight when the requirement is that nothing leaves the machine and the consumer is an agent, not a human IDE.

Agent relevance

API CLI MCP Behavioral-testable

Agentic-Commerce Readiness 75/100 · COMMERCE-READY

Independent readiness for agent delegation & transaction. How it’s scored · check live

Direct MCP server (srclight serve) with 25 tools, connectable from Claude Code, Cursor, Windsurf or any MCP client. Also a CLI for indexing. Purpose-built to be driven by agents for code retrieval — highly composable.

Agent-friendly score: 9/10

Evidence

Deep code index for AI agents: SQLite FTS5 + tree-sitter + embeddings + MCP, 25 tools — source (2026-09-01) verified
Fully offline — local embeddings via Ollama; single SQLite file, no external DB/Docker/cloud — source (2026-09-01) verified
One-line install (pip install srclight); MIT license; multi-repo workspaces — source (2026-09-01) verified

scorecard.json · transparency passport · registry · methodology

More: compare agents · best of · developer tools · incident registry

Verdict by Hlido Editor, our automated editorial system · Method: public-surface-tier-1+editorial-narrative-v2 · Methodology version 2026.05 ·

How this page was produced. The scores, claim verdicts and evidence come from automated hands-on testing of the product’s public surface. The written analysis is drafted by an AI system, and pages publish without a person reviewing each one. Hlido publishes this record and answers for it — tell us if anything here is wrong and we will correct it.

Embed this trust badge

Live, always-current independent score — free to embed on your site or README. No vendor pays for placement.

Markdown

[![Hlido trust score](https://hlido.eu/badge/srclight-srclight.svg)](https://hlido.eu/check/?agent=srclight-srclight)

HTML

<a href="https://hlido.eu/check/?agent=srclight-srclight"><img src="https://hlido.eu/badge/srclight-srclight.svg" alt="Hlido trust score"></a>