Omi CLI · v0.1

Omi can build anything. Now from your terminal.

An AI coding agent that works inside your repo: it reads the code, plans, edits files with diffs you approve, runs your commands and tests, and fixes failures until they pass.

  • macOS, Linux, Windows
  • Node.js 18+, no other installs
  • Start free, prices in USD
install
$ npm install -g @builditomi/cli

Then open a project and type omi. The first run signs you in with your BuildItOmi account in the browser.

What it does

A senior engineer in your shell

All commands
  • Reads your codebase

    Lists, searches (ripgrep when you have it) and reads files to understand the code before it changes anything.

    list · glob · grep · read
  • Plans first

    Shows a live step-by-step plan. /plan makes it read and plan only; /build applies the plan.

    /plan · /build
  • Edits with diffs you approve

    Every edit is shown as a coloured diff. Say yes, no, always, or type what to do instead.

    edit · write · delete
  • Runs commands, with permission

    Builds, linters, package managers, git. Read-only commands run; anything else asks first.

    run_command
  • Tests and fixes in a loop

    /test finds your test command, runs it, fixes the code and runs it again until the tests pass.

    /test
  • Git-aware

    Knows your branch, changes and recent commits. /diff, /branch, and /commit with a message it writes and you approve.

    /diff · /commit · /branch
  • Sessions you can resume

    Saved after every step. Pick up where you left off, even after closing the terminal.

    omi --continue · /resume
  • Project memory

    OMI.md holds your project's commands and conventions, read at the start of every session.

    /init · /memory
  • Helpers that work in parallel

    Research questions go to explore helpers that read the code on their own, several at once, and report back with file:line references.

    explore
  • Reads the web when it needs to

    Fetches a link or current docs, and searches Google with source links when it needs facts it doesn't have. Both ask first.

    web_fetch · web_search
  • Connects your tools with MCP

    Add any stdio MCP server, like your issue tracker or database, and the agent can use its tools. Each call asks first.

    omi mcp add
  • Files and screenshots in a prompt

    Mention @file to attach it, or attach a screenshot and Omi sees it. Answers stream in as they are written.

    @path

Control

You decide what changes

  • Four modes

    ask (default) asks before edits and commands; auto-edit applies edits and asks before commands; yolo runs everything; plan changes nothing.

  • Undo a whole task

    /undo puts back every file the last task changed, even without git.

  • Stays in your project

    File tools refuse paths outside the project folder. Commands run in your shell on your computer, after you approve them; never on our servers.

  • Your code is not stored

    What Omi reads is sent only to run the model for that turn. We keep token counts for billing, not your code or conversations. Sessions live in ~/.omi on your computer.

Pricing

Pay for the model work you use

All plans

Start free

$0

Every BuildItOmi account gets 1,00,00,000 free Omi CLI credits every month to use on real code. No card.

Plans

From $20/mo

Pro $20, Pro+ $60 and Max $200 a month, with 12,00,00,000 to 1,20,00,00,000 credits every month. Yearly gets 2 months free.

Every turn

Real cost

Each agent turn is charged the tokens it used. You see the credits after every task. Packs of $10 and $25 never expire.

Start in a minute

Three commands

How it compares
  1. Install

    One package, no other installs.

    npm install -g @builditomi/cli
  2. Sign in

    Approve a code in your browser.

    omi login
  3. Give it a task

    In any project folder.

    omi "fix the failing tests"