Free

locus index and locus serve run entirely on your own machine (macOS or Linux; Windows is not yet verified). With LOCUS_EMBEDDER=local (the default), embeddings are computed on-device with sentence-transformers — no API key, no account, no signup, and nothing sent to a server we operate. When your AI assistant searches, the matching excerpts go to that assistant. This is not a limited trial or a teaser tier: it has no time limit and no token cap, and it's the mode this project is built to keep working indefinitely (see the local-first constraint in the project's own engineering docs).

The limits that do apply here are architectural, not billing gates, and they're the same regardless of which mode you run: a default 25 MB per-file cap (LOCUS_MAX_FILE_MB, configurable), and a fixed set of file formats the parser actually understands — plain text and code files, Markdown, PDF, and .docx. Anything outside that set is skipped, not partially indexed. The full breakdown, pulled from the real parser code, is on the compatibility page.

Search quality depends on which embedding model is doing the work — the on-device model is smaller than the hosted ones — but nothing in the search/retrieval path itself is held back behind a paywall in this mode. The free local mode is optimized for English; managed mode uses a multilingual model. We have not yet measured search quality in other languages. Setup steps: connect it to your AI assistant.

Creating an account gets you the managed embeddings relay: you don't need your own Voyage API key. On managed plans, chunk text passes through Locus's relay to Voyage AI to be embedded. We don't store it; we keep only usage counts. This is a trial, not a subscription tier, with a 5,000,000-token budget for the whole trial. No credit card is required to start it.

How long it lasts. While paid plans are not open yet, a free trial does not run out by time. Once they open, it runs until 7 days after they open or 14 days after you signed up, whichever is later. Usage limits still apply: the trial's token budget, a daily limit, and at busy times a pause on trial indexing until the next UTC day.

Once the trial ends — by time or by its token budget — continuing on the hosted relay means picking a paid plan, which you can do once paid plans are open. See pricing for the current plans; none of that is required to keep using the local engine above, which the trial doesn't affect.

There is no permanent free hosted tier today. A small ongoing hosted allowance — something like a modest monthly amount of managed embeddings at no cost, separate from the trial above — has been discussed internally as a possible future addition. It is a proposal under cost/benchmark evaluation, not a shipped feature, and nothing on this site should be read as already offering it. If it's built, this page gets updated to describe what actually exists then — not before.

• Comfortable with a terminal, want it free and running forever with no account: install locally (macOS or Linux; Windows is not yet verified) and use LOCUS_EMBEDDER=local.
• Want managed embeddings without generating your own API key, and are fine picking a paid plan once the trial ends: start the hosted trial from signup.
• Want the exact current prices and daily quotas before deciding: see pricing.