Three products. One ground truth

Ground truth for your agent, review for your PRs, root cause for your incidents. Each priced on its own, simply.

Niro Local

Ground truth for your agent, built and stored on your machine. Graph traversal, not an LLM call.

FreeLimited time
$0no card needed

For your own projects and for trying Niro on real code.

  • Up to 500,000 lines of codeNo line limit when you turn on code sharing
  • Call chains, impact analysis and cross-repo links
  • Answers in the assistant you already use
  • Graph built and stored on your machine
Get started free
Pro
$20/month

For codebases past the free line limit, without sharing any code.

  • Everything in Free, with no line limit
  • Nothing shared unless you choose to
  • We reply within one business day
Request free trial
Teams and enterprise
Contact us

For organisations

For teams that need a data processing agreement and admin controls.

  • Everything in Pro
  • Data processing agreement and admin controls
  • 30-day trial key within one business day
Contact us

What leaves your machine

Code sharingOff by default

Turn it on to lift the Free line limit. Niro extracts structural signals to improve graph accuracy. Code is deleted right after processing and never used to train generative AI.

Coverage findings and tool usageOn every plan

A few redacted files the graph could not resolve, so we can fix the parser, and which Niro tools your assistant called. Your graph itself stays on your machine.

Niro PR Review

Real-time review, grounded in structure. Cross-service dependencies and blast radius, before merge. Codebase chat and graph visualisation included.

PR Review
$25/seat/month

30 reviews/seat/month included, then $1/review

One shared graph, every PR checked against it before merge.

  • 30 PR reviews/seat/month, overage $1/review
  • Codebase chat, 1,000 messages/seat/month includedCalls a model, so it's metered like reviews, not left unlimited
  • Interactive graph visualisation, includedNot a model call, no limit
  • Resolved edges, like a compiler, not inferred from similarity
Start free trial
Enterprise
Custom

Unlimited reviews, SSO, SLA

For organisations that need unlimited volume and compliance.

  • Unlimited PR reviews
  • Unlimited codebase chat
  • SSO, audit trail, dedicated support

Niro RCA

From incident to root cause, in minutes. Automated graph traversal fires on alert. No payment decisions during an incident, we bill after resolution.

Enterprise
Custom

Unlimited, org-wide, in contract

For teams with frequent incidents, no per-incident billing.

  • Unlimited incidents, no per-incident charge
  • Org-wide scope, every indexed repo
  • Dedicated support

How this actually works

Each product stands alone

Niro Local, PR Review, and RCA are priced and billed separately. Use one, two, or all three, there's no bundle requirement and no cross-product discount built in yet.

Why the free line is a line count

500,000 lines is something we can measure on your machine, not something you have to declare. Each project counts once, by its largest folder, so two checkouts of one repo are not counted twice.

Sharing is a choice, never a requirement

Free works with sharing off, up to the line limit. Sharing lifts the limit; Pro lifts it with nothing shared. Shared code is never used to train generative AI.

Why PR Review has a cap

Unlimited flat-rate review pricing breaks down under real agentic PR volume. A included allowance plus a small overage keeps the price simple without an open-ended cost risk on either side.

RCA billing during an incident

We never hard-stop mid-investigation. RCA always completes regardless of billing status, we invoice after the incident is resolved, not before or during.

Enterprise across all three

Volume pricing, SSO, audit trail, and dedicated support are available per product or as one combined contract. Talk to us once, we'll scope whichever mix you need.

FAQ

What is ground truth?

Facts about your code that come from the code itself, not from a model's guess. Niro parses your repositories into one graph and resolves every call, import and endpoint the way a compiler does, so a question like "what calls this function" is answered by a lookup with a file and line behind it. Your agent, your reviewers and on-call read the same graph, so they get the same answer.