Autonomous AI agent · runs on your machine

The AI that stops advisingand ships code

NoTrack Coder runs real tools from one chat — shell, files, SSH, databases, the browser, image generation and web search. It doesn't hand you snippets. It executes.

Uncensored, model-agnostic, and 100% local — no account, no telemetry.

Download for Windows

Windows 10 / 11 · x64 · installs in about a minute

The whole point

Every other assistant stops at advice.
This one closes the loop.

Most tools hand you a suggestion to copy-paste. NoTrack Coder runs the command, edits the file, drives SSH, databases and the browser — on your machine, end to end — and shows every action as a card you can expand. From a plain-language request to a shipped change.

What you get

Everything runs from one chat.

Acts, not advises

Runs shell / PowerShell, reads & writes files, SSHes into servers, queries MySQL / Postgres — real results, not snippets.

Any model, or offline

Any OpenAI- or Anthropic-compatible endpoint, or a local model via Ollama, LM Studio, llama.cpp — bring your own or run fully offline.

100% local, no telemetry

Chats, working folders and API keys never leave your machine. Secrets in tool output are redacted before they reach the model.

Connectors (MCP)

Plug in Slack, databases, the browser (Playwright / Puppeteer / Chrome) and more over one standard interface — by command or URL.

Vision & media

Built-in image generation, vision and Windows OCR — plus web search that queries two sources in parallel.

Automation

Parallel sub-agents, scheduled cron-agents and reusable skills. Fire off work and let it run on its own.

How it works

From a sentence to a shipped change.

1

You ask

Describe the outcome in plain language.

2

Agent plans

It breaks the job into steps and picks the tool each one needs.

3

Tools run

Shell, files, SSH, database, browser — executed on your machine.

4

You review

Every action and its result lands in the chat. Approve or adjust.

Pricing

Same agent. You pick the speed.

Every tier runs 100% locally with no telemetry. Free to try, no account needed to download.

StandardThe regular version
$199 / month
The regular build. Full capability at standard speed — not the fastest, but no compromises.
  • Regular speed
  • Unfiltered model
  • All tools, connectors & automation
  • Web search, image generation & vision
  • 100% local — no telemetry
  • Auto-updates
Most capable
ProAs we run it
$499 / month
The exact build we use ourselves. Maximum speed, no queue, no delay, nothing held back.
  • Maximum speed — no queue, no delay
  • Extra text model — notrack-chat in the base set
  • Everything in Standard
  • Run local models — Hugging Face & any installed locally
  • Priority auto-updates
  • Prompt help for your project — via Support

Prices in USD. All tiers are the same local agent — the difference is speed, the model, and a few Pro extras. The free tier (no cost, queued speed, filtered model) is available from the download above.

FAQ

Questions & answers.

What is NoTrack Coder?+
An autonomous AI coding & infrastructure agent for Windows. It doesn't just advise — it runs real tools from one chat: shell / PowerShell, files, SSH, databases, the browser, image generation, vision and web search.
Is my data private?+
Yes. It runs 100% locally on your machine — chats, working folders and API keys stay on your computer. There is no telemetry and no account is required to run it. Secrets that show up in tool output are redacted before anything reaches the model.
Which models can I use?+
Any OpenAI-compatible /v1 endpoint (OpenAI, vLLM, llama.cpp, LM Studio, Ollama, OpenRouter, proxies), the Anthropic Messages API, or a local ComfyUI for images. A built-in uncensored coding model is included out of the box, and you can run open models — including from Hugging Face — locally.
Is it really uncensored?+
The built-in model answers directly — no fake refusals, no moralizing. You stay in control with per-chat approval modes: Auto-run, Ask first, or Plan only.
What are the system requirements?+
Windows 10 / 11 (x64). The installer sets up the WebView2 runtime automatically and takes about a minute.
How much does it cost?+
Free to download and try. Paid tiers — Standard ($199/mo) and Pro ($499/mo) — differ in speed, model filtering and a few Pro extras like extra local models and prompt help. See pricing above.

A local, uncensored AI coding agent for Windows

NoTrack Coder is an autonomous AI coding agent for Windows that acts instead of only advising. From a single chat it runs shell and PowerShell commands, reads and writes files, SSHes into servers, queries MySQL and Postgres databases, drives the browser, generates images and searches the web — then loops until the task is done, showing every action and its result inline. It closes the loop from a plain-language request to a shipped change.

Private and local by default

Everything runs on your machine. Your chats, working folders and API keys never leave your computer, and there is no telemetry. Secrets from your .env files, bearer tokens and private keys are automatically redacted from tool output before they ever reach the model — so a no-track, local-first AI assistant keeps your source code and prompts off other people's servers.

Bring any model, or run fully offline

There is no model lock-in. Connect any OpenAI- or Anthropic-compatible endpoint, or point NoTrack Coder at a local runtime such as Ollama, LM Studio or llama.cpp — including open-weight models from Hugging Face — to run entirely offline. Prefer the built-in uncensored model, a hosted frontier model, or a small local one for quick tasks, and switch whenever you like.

Automate real work

Connect MCP connectors for Slack, databases and the browser through one standard interface, fan work out to parallel sub-agents, schedule cron-agents that run on their own, and keep reusable skills and persistent memory between chats.

Get started

Stop copy-pasting.
Start shipping.

Free to download · Windows 10 / 11 · runs entirely on your machine.

Download for Windows