Newmiii now works in your browser too — a desktop-style app. Run miii webSee it ›
🇮🇳 Made in IndiaThe open alternative to Claude Code ›

Claude Code's workflow.
Every model. Free forever.

A professional-grade coding agent for your terminal or browser. It plans, edits, runs and verifies — with Claude, GPT, Gemini, DeepSeek, or a model running 100% on your own GPU.

Compare with Claude Code
miii — ~/code/my-app
>⏎

One agent. Every model.

Head to head

Everything you like about Claude Code.
None of the lock-in.

Claude Code is excellent — if one vendor, one model family and one bill suit you. miii gives you the same agent loop and lets you choose the rest.

FeaturemiiiClaude Code
The tool itselfFree, MIT-licensedPaid plan or API billing
Models14+ providers — Claude includedClaude only
Local models on your GPUOllama, LM Studio, llama.cpp, vLLMNo
Works with the wifi offYes, with a local modelNo
Where your code goesWherever you choose — or nowhereAnthropic or your cloud
Source codeOpen sourceProprietary
Switch model mid-sessionAny provider, /providerBetween Claude models
Plan mode, subagents, MCP, hooksYesYes
Rewind files + conversation/rewindYes
Browser / desktop appmiii web, on localhostYes
Fix-until-green commandmiii fix, with auto-rollbackNot built in
Grade models on your hardwaremiii doctor—
Claude Code details from Anthropic's public docs, October 2026. Plans and features change — check theirs before deciding.

Switching

Your muscle memory
already works here.

Same hook events, same flags, same slash commands. Moving over takes a minute — and you can keep using Claude through your own API key while you try other models.

Plan mode

/plan

Read-only research, then you approve the plan.

CLAUDE.md

MIII.md

/init writes it, # appends a line, /memory shows where it lives.

Subagents

.miii/agents/*.md

Project or user scope, called through the task tool.

Custom commands

.miii/commands/*.md

Markdown files become slash commands.

Hooks

PreToolUse · Stop

Same events: PreToolUse, PostToolUse, UserPromptSubmit, Stop.

Headless

miii -p

--output-format json, --allowed-tools, --max-turns.

Permission modes

acceptEdits · plan

default, plan, acceptEdits, bypass — rules saved per project.

Checkpoints

/rewind

Puts the files and the conversation back together.

Context tools

/compact /context

Plus /cost, /review, /export, /sessions, /vim.

Keep Claude. Lose the lock-in.

Three lines to move over.

miii picks up the ANTHROPIC_API_KEY you already export. When you want to try something else, /provider switches backends without restarting the session.

npm i -g miii-agent
miii provider add anthropic
# uses your ANTHROPIC_API_KEY
miii
  • Anthropic
  • OpenAI
  • Gemini
  • Groq
  • OpenRouter
  • DeepSeek
  • Mistral
  • Together
  • xAI
  • Cerebras
  • Ollama
  • LM Studio
  • llama.cpp
  • vLLM

Only in miii

Most AI tools rent you a cloud.
miii lets you choose.

Use the best hosted models for hard problems, free tiers for experiments, or a model on your own GPU — where nothing is metered, nothing is uploaded, and it keeps working with the wifi off.

Private when you want it

Go local, and your code never leaves your machine.

Cloud agents ask you to trust your codebase to someone else's server. Point miii at a local model and it only ever talks to localhost — no telemetry, no analytics.

your repo→localhost:11434→your GPU

Cost

$0

No paid tier, ever. Local models cost nothing to run.

Offline

Yes.

Air-gapped repos, NDA'd code, a flight with no wifi.

Fix until green

The tests decide what's fixed. Not the model.

miii fixruns your check, fixes what fails and runs it again. Attempts that don't help are rolled back, so even a small local model only moves the code forward.

miii fix --check "npx tsc --noEmit"--output-format json

Decision box

A second model checks the work is really done.

Before the agent stops, a smaller model verifies the task is actually complete — fewer "done!" replies that aren't.

planeditrunverify✓ done

Model doctor

Know which model works.

miii doctor grades your installed models on real engineering tasks.

Model-aware

Small models, real tool calls.

miii repairs broken tool calls and tunes prompts to each model's context window.

Permission-gated

You approve every change.

Writes and shell commands ask first. File tools stay inside your workspace.

Quick start

Up and running in a minute.

Requires Node.js 18 or later. For local models, run Ollama; for hosted ones, miii provider add <name> picks up the key you already export.

1

Pick a model

ollama pull qwen2.5-coder:14b
2

Install miii

npm install -g miii-agent
3

Talk to it

miii
> @src/server.ts add rate limiting
> why are my tests failing?

Web app

Rather have a window?
miii web

The same agent in a desktop-style app in your browser — for when you'd rather click than type. Same sessions, same models, same permission rules as the terminal.

Chat, diffs and approvals

Everything the terminal shows, laid out like a chat app.

Replies stream in, each tool call opens to show the exact command and its output, edits show as real diffs, and every write waits for your click. Paste images, and pick model, provider, effort and mode from the input bar.

One command

Open it from any project.

It starts a local server and opens your browser. Add --port, --no-open or -c to resume your last chat.

miii web
# or: miii --web

Locked to your machine

Yours alone.

It listens on localhost only, and every request needs the one-time key in the link miii prints. Other sites can't reach it.

127.0.0.1#token=…

Commands

Everything is a / away.

The same command list you get inside miii. Type to filter, then press enter or click to copy. Press / anywhere on this page to jump here.

    ↑↓ navigate · ⏎ copy · esc clear

    Shipping fast

    What's new.

    Read live from CHANGELOG.md every time the page loads.

    Loading releases…

    Full changelog

    Questions.

    Can I keep using Claude models?

    Yes. Run miii provider add anthropic and miii uses the ANTHROPIC_API_KEYyou already export. You pay Anthropic's API rates for what you use, and you can switch to another provider or a local model at any point with /provider.

    Is miii really free?

    Yes. miii is MIT-licensed with no paid tier, no account and no sign-up. What you pay for is the model: nothing if it runs on your machine, and whatever your provider charges if you use a hosted one — Gemini, Groq, OpenRouter and Cerebras all have free tiers.

    Is my code ever uploaded?

    With a local provider, your code only ever travels to localhost. No telemetry and no analytics — the one outbound call miii makes on its own is a version check against registry.npmjs.org, which carries nothing but the package name. Add a hosted provider and that provider sees what you send it, which is why local is the default.

    Can it break my repo?

    Every write and every shell command asks first, and file tools are confined to the working directory. When something does go wrong, /rewind puts the files and the conversation back.

    Do I need a GPU?

    No. Use a hosted provider and any laptop will do. For fully local models you want one — qwen2.5-coder:3b runs on CPU, but it will be slow and need more hand-holding. Apple Silicon works well: unified memory counts as VRAM here.

    Why not just use Claude Code?

    Use it if it fits. miii is for when you want the same workflow without being tied to one vendor or one subscription: pick any model, pay only for what that model costs, or go fully offline for air-gapped work, NDA'd code, or a flight with no wifi.

    Own your tools.

    No account, no subscription, no lock-in. A coding agent in your terminal in under a minute.

    ★ Star on GitHub

    Apna code, apna model, apni marzi.