grep finds strings.
grepvec finds meaning.

Code intelligence for developers and AI agents. Structural search, graph context, precise source extraction. Your agent's context window stays clean.

curl -fsSL https://grepvec.com/install.sh | sh

Every grep match, every wrong file opened, every irrelevant function — that's context window pollution. Your agent reasons over noise. Its decisions degrade.

grepvec eliminates this.

The research loop

Three commands. Search, understand, read. No noise.

$ grepvec search "how does authentication work" ━━━ Pass 1: Keyword Search ━━━ 1. function api::hmac_auth::validate_signature (src/api/hmac_auth.rs) — 42 LOC Called by: authenticate_request, verify_lab_signature ━━━ Pass 2: Neural Search ━━━ 1. [neural] api::hmac_auth::HmacAuthError — score 0.52 enum with variants: InvalidTimestamp, InvalidSignature, MissingHeader
$ grepvec context "validate_signature" Calls (3): → hmac::Mac::verify (unresolved) → hex::decode (unresolved) → chrono::Utc::now (unresolved) Called by (2): ← authenticate_request (src/api/routes.rs:7) ← verify_lab_signature (src/api/rest.rs:35)
$ grepvec read "validate_signature" function api::hmac_auth::validate_signature (src/api/hmac_auth.rs:45–86) 45 pub fn validate_signature( 46 secret: &[u8], 47 timestamp: &str, 48 body: &[u8], 49 signature: &str, 50 ) -> Result<(), HmacAuthError> { .. ...

How it works

01

Parse

tree-sitter extracts every function, struct, trait, and their relationships from your codebase. Deterministic. Reproducible.

02

Map

Callers, callees, dependencies stored in Postgres. A structural graph of your entire codebase — not text, but meaning.

03

Search

Keyword (tsvector) + neural (vector embeddings). Natural language queries return precise, structurally-ranked results.

Agent integration

grepvec ships with MCP server and subagent instructions. Your AI tools discover it automatically.

Claude Code, Cursor, Codex — one grepvec init and your agents have code intelligence.

Claude Code Cursor Codex MCP Subagent

Vector backend

Neural search coming soon. Keyword search works today.

managed

Enscribe

Managed vector storage, embeddings, and neural search. Zero infrastructure.

self-hosted

Local

BGE embeddings + Qdrant. Run everything on your machine. Full control.