Skip to main content

Reason, don't vector.

Knowing by reasoning, not vectors.

Deep and reliable. Vectorless plays nicely with your documents. Ask questions in plain language; get answers by reasoning.

Core Engine

vectorless

A reasoning-based document understanding engine for AI. Compile documents into a rich IR, query with an agent that navigates and reasons. Zero embedding dependency.

For AI engineers building retrieval systems.

pip install vectorless
GitHub
Application

vectorless-code

AI code search for your entire codebase. CLI + MCP server that plugs into Cursor, Claude Code, or any AI coding tool. No vector DB, no embedding model — just compile and search.

For developers who search code every day.

pip install vectorless-code
GitHub

How it works

1
Compile. Parse your documents (or codebase) into a rich intermediate representation — a navigable tree with keyword indexes, routing tables, and evidence scores baked in. No LLM required.
2
Reason. An AI agent navigates the tree like a human expert —ls to explore, cd to dive deeper,cat to read, find to search. It reasons about which path leads to the answer.
3
Answer. The agent collects evidence with full source attribution — section title, node path, line numbers. Every claim is traceable.
GitHubGet Started