Find code by describing
what it does.

Name: Savants
Author: Savants

Not what it's called. 90% accuracy. Under 400ms. Offline.

curl -fsSL savants.sh | sh Copied!

Works on macOS and Linux. No API keys. No cloud required.

Scroll

grep

$ grep -rn "validate" src/

src/config.js:14: // validate config

src/utils.js:87: validateEmail(input)

src/auth.js:23: // TODO: validate token

src/form.js:91: if (!validate(form)) {

src/logger.js:45: validateOptions(opts)

... 9 more matches

14 matches. None are the one you need.

grep searches text.
savants searches meaning.

savants

$ savants search "function that validates log level options"

1 result (387ms):

lib/logger.js:validateLogLevelOption

Lines 142-178 | Score: 0.94

Validates the log level passed via CLI flags,

normalizes aliases, returns canonical level.

90%

accuracy

<400ms

cached query

28x

compression

API keys required

Tested on Fastify (33K stars, 287 files, 2,152 functions)

Three commands. That's it.

Step 01

Install

One line. No dependencies. No API keys. Runs entirely on your machine.

curl -fsSL savants.sh | sh

Step 02

Index

Parses every function with tree-sitter, embeds with ONNX. Your code never leaves your machine.

savants reindex

Step 03

Ask

Your AI agent calls savants instead of grep. Semantic search, call graphs, file skeletons - all local.

savants search "auth middleware"

38 tools. 5 you'll use daily.

Every tool is local, fast, and designed for LLM agents.

semantic_search Find code by concept ~400ms

file_skeleton 28x fewer tokens ~85ms

callers Every caller, confirmed ~15ms

where_used All usages, all files ~15ms

reindex Auto-detects stale indexes auto

141 seconds vs 17 minutes.

We diagnosed a real Fastify bug with savants. Identified root cause, traced the call chain, and submitted PR #6683 - all in under 3 minutes.

Read the case study →

Local tools are free. Forever.

Cloud tools add production intelligence. Pay per call, no commitments.

diagnose-error $5.00

Root cause file + line in 0.7s. Upstream trace, git blame, Slack context.

pr-risk $2.00

8-check risk analysis per PR. Blast radius, affected consumers, test gaps.

diff-impact $1.00

Full downstream impact analysis across services and modules.

radar $1.00

What-did-I-miss digest. Knows when your graph drifts from production.

10 free cloud calls/month. No card required. Start free →

$ savants connect

Find code by describingwhat it does.