Indexed and connected
in about a minute
One command, nothing to install. Signing in, pointing it at a folder and connecting your assistant are prompts it walks you through.
Sign-in, credentials and branches
The boundaries matter more than the features, because they are what you are trusting.
Ask your agent whether you are logged in and it runs niro whoami. If you are not, it asks you to sign in yourself. It cannot drive a browser sign-in and does not handle your credentials.
A long-lived credential is written to ~/.niro/credentials.json at mode 0600. Short-lived tokens refresh from it.
Every question routes to the project covering your repo and current branch. On a branch no project covers, Niro refuses to answer from stale code and offers to build a temporary project instead.
Two worktrees of one repo are read as two trees, keyed separately, sharing only what is safe to share.
| Before | After one index | |
|---|---|---|
| Finding a caller | grep, then read | one call |
| Cross-repo calls | only what is checked out | every indexed repo |
| Config values | a file it may not open | nodes on the graph |
| Stale branches | answers anyway | refuses, and says so |
FAQ
What do I need installed?
Node 18 or newer. Then npx @niroai/niro@latest start, which signs you in through the browser with GitHub or Google, creates your account on the first run, installs the local engine, builds the graph for the folder and connects your assistant.
I use more than one assistant
Run the connect step again for each one. It wires a single client per run, thirteen clients are supported, and every one of them reaches the same graph with the same 35 tools.
Where do my credentials live?
A long-lived credential is written to ~/.niro/credentials.json at mode 0600 on your own disk, and short-lived tokens refresh from it. The CLI cannot drive a browser sign-in and does not handle your credentials on your agent's behalf: ask your agent whether you are signed in and it runs niro whoami, then asks you to sign in yourself.
Does it work inside a git worktree?
Yes. Two worktrees of one repository are read as two trees, keyed separately, sharing only what is safe to share. On a branch no project covers Niro refuses to answer from stale code and offers to build a temporary project instead.
What happens if Niro does not work?
Your assistant keeps working. When Niro cannot answer, the assistant falls back to its own tools, reading files and searching, as it would without Niro. Run niro status to check the folder’s graph, niro build to rebuild it, or niro start to run the setup again. If it still fails, write to ask@niro.ai.
Which languages are supported?
Java, Kotlin, JavaScript (including JSX, Vue and Svelte), TypeScript (including TSX and Astro), Python, C#, Go, PHP, Ruby, Swift, Rust, C and C++. One repo can mix them, and every plan supports all of them.
What does the Niro CLI share?
Your code stays on this machine, except, for each line the audit flags, the whole method that holds it and one line for each name it uses, sent with secrets masked and author and company names removed. A record of which Niro tools your assistant used is also sent, to improve Niro and to show you its value. Niro’s tools answer your assistant from the graph on this machine and never send the contents of your files. Help improve Niro is off by default on every plan; when you turn it on, Niro redacts your code on your machine first (secrets, author names and company names are removed), extracts structural signals from it to improve graph accuracy, and deletes it right after processing. It is never used to train generative AI.
Can you add support for my language?
Yes, ask us. Write to ask@niro.ai with the language, and the frameworks you use with it.
Index one repo and ask it something
Start with the service you know least well. Ask your assistant how it works, and compare the answer with what it would have said yesterday.
Full command reference and options: npmjs.com/package/@niroai/niro