Skip to main content

Your files. A deeper you.

Give your AI your kind of context.

Search your own files from the AI assistant you already use — with the source file for every answer.

Your raw files are never uploaded to Locus. When your AI assistant searches, the matching excerpts go to that assistant, as they would if you pasted them in. How each mode works

Locus works with Claude Code, GitHub Copilot CLI, Codex, Claude.ai (web app) and ChatGPT (general chat, not Codex), with cited answers. Claude.ai and ChatGPT connect through Locus's hosted relay.

What Locus does, in three steps, on a demo folder

01 · Your filesDemo folder, fictional company

Point Locus at folders on your own machine.

~/LocusDemo/demo-vault/21 files

  • planning/launch-plan.md
  • planning/risks.md
  • marketing/press-kit-draft.md
  • pricing/pricing-v3.md
  • meetings/2026-08-19-pricing-review.md
  • …16 more

Your raw files are never uploaded to Locus.

02 · Relevant passagesDemo folder, fictional company

Your assistant asks; Locus returns the matching passages with their paths.

YOU

Which two documents disagree about the launch date, and what does each say?

assistant → locus_search · two of the matching passages, shortened

  • planning/launch-plan.mdchunk 0

    “Public launch is Tuesday 3 November 2026. App-store submissions go in two weeks earlier…”

  • marketing/press-kit-draft.mdchunk 0

    “…launches publicly on 17 November 2026 on iOS, Android and the web.”

When your AI assistant searches, the matching excerpts go to that assistant, as they would if you pasted them in.

03 · Cited answerDemo folder, fictional company

Your assistant answers and points back to the files it used.

They disagree. The launch plan says Tuesday 3 November 2026source 1; the press kit draft says 17 November 2026source 2.

  1. planning/launch-plan.md· chunk 0
  2. marketing/press-kit-draft.md· chunk 0

Your assistant answers it and points back to the exact files and passages it used. Answers vary by assistant and model; this one follows the demo folder's answer key.

You have years of files and notes you can never quite find.

Locus makes them searchable from the AI assistants you use, with the file path and passage behind every answer.

Paths from the demo folder (fictional company).

What Locus works with

One index on your computer.
The assistants you use.

No accuracy scores here yet: our early tests ran on one personal vault, too small to generalise from.

Your computer

Locus: your folders, indexed

MCP tools: locus_search, locus_read_file, locus_list_sources

  • Claude Code
  • GitHub Copilot CLI
  • Codex

The connect guide labels every other client by how far along it is

5

AI clients Locus works with, with cited answers: Claude Code, GitHub Copilot CLI, Codex, Claude.ai (web app) and ChatGPT (general chat, not Codex). The connect guide labels every other client by how far along it is.

Path + passage

on every search result, so your assistant can cite the exact file it used

macOS · Linux

where Locus runs today; Windows is not yet verified

How it works

  1. Index your files locally

    Point Locus at folders on your own machine. It parses, chunks, and embeds your files into a local vector index. Your raw files are never uploaded to Locus. On managed plans, chunk text and your search queries pass through Locus's relay to Voyage AI to be embedded. We don't store them; we keep only usage counts. The free local mode is optimized for English; managed mode uses a multilingual model.

    ~/LocusDemo/demo-vault · 21 files
  2. Connect the AI assistants you use

    Locus speaks MCP, the open standard for connecting tools to AI models. Locus works with Claude Code, GitHub Copilot CLI, Codex, Claude.ai (web app) and ChatGPT (general chat, not Codex), with cited answers. Claude.ai and ChatGPT connect through Locus's hosted relay. The connect guide has setup steps for more, each labelled with how far along it is. See the full list

    locus_search
  3. Get cited answers

    Ask a question that spans years of notes and documents. Your assistant answers it and points back to the exact files and passages it used.

    planning/launch-plan.md · chunk 0

Pricing

The core engine is free and runs on your own machine. Paid tiers add the hosted extras.

Starter

$4/mo

$48 billed yearly + applicable taxes

One computer: your notes and documents, searchable from your AI assistant.

Enough for a typical folder of notes and documents

Join the waitlist

Personal

$8/mo

$96 billed yearly + applicable taxes

Up to three computers, with room for larger document collections.

Room for larger collections and regular re-indexing

Join the waitlist

Power

$16/mo

$192 billed yearly + applicable taxes

Up to ten computers and the highest daily indexing limit.

The highest limits, for big archives and code

Join the waitlist

Paid plans are not open yet; waitlist members get an email when Locus opens; no launch offer is promised until its terms are published.

Free: the local engine · See full pricing

Join the waitlist

Paid plans are not open yet. New accounts open with them. Leave your email and we'll write once, when Locus opens: waitlist members get an email when Locus opens; no launch offer is promised until its terms are published.