Browser and Computer Use
DCode drives a built-in browser to check the pages it builds, and when a workflow leaves the terminal it can see the screen and operate native desktop apps. Both are ready the moment you install.
Describe a task in plain language. DCode plans it, splits it across up to 24 AI agents working in parallel, edits your files, runs your build and tests, and tells you exactly what it verified.
curl -fsSL https://trydcode.app/install.sh | sh
The Crew
Type the task the way you would say it to a teammate. There are no tickets to split and no agents to configure.
A crew lead reads your repository, breaks the task into independent parts and writes a shared brief, so names, interfaces and design choices stay consistent across the crew.
Each agent takes a part. The Crew dock shows who is doing what and who needs you. Answer a question once and every agent asking it gets the answer.
A reviewer runs your real build and tests, then merges the parts. You get one answer that says exactly what was verified.
Agentic by default
DCode plans multi-step changes, calls tools in sequence and keeps iterating until the issue is resolved. It reads the error output from your compiler, linter and test runner, fixes what fails and runs them again.
You stay in charge
Choose how much freedom DCode gets, from Plan mode to Full access. In Auto, DCode checks each step, runs the safe ones and asks before anything risky, such as pushing to git, deploying or touching credentials.
From task to pull request
Run a task in its own worktree so it never touches your working copy. When the work is ready, DCode pushes the branch, opens the pull request and shows its checks. If one fails, Fix failing checks starts a task that finds the problem and pushes the fix.
Beyond the editor
DCode works where the job actually happens: in the browser, on the desktop, in your databases and across long jobs that run for hours.
DCode drives a built-in browser to check the pages it builds, and when a workflow leaves the terminal it can see the screen and operate native desktop apps. Both are ready the moment you install.
Connect any Model Context Protocol server in one click. Query databases, read issues, inspect live API documentation and call your own internal tools from the same conversation.
The official marketplace installs plugins and skills in one click, hash-verified on every download. Built-in plugins create Word documents, PDFs, slide decks and spreadsheets, and you can write your own.
For long, structured jobs, DCode designs a workflow of phases and agents, shows it as a live timeline, and lets you pause, resume or amend a run.
Schedule a task and DCode runs it on its own. Scheduled runs use Auto, so risky steps are refused instead of done.
Raven's thinking streams render in plain view, so you can follow why a change was made before you accept it.
DCode CLI
dcode opens a full-screen terminal workspace on Windows, macOS and Linux. Same account, same agents and same Raven models as the desktop app, plus flags built for scripts and CI.
| Command | What it does |
|---|---|
dcode | Open the terminal app in the current folder |
dcode -p "fix the failing tests" | Run one task and print the result |
dcode -p "..." --crew always | Run the task with a crew |
dcode --worktree | Work in a separate worktree |
dcode mcp | Add, list and remove MCP servers |
dcode schedule | Run a task on a schedule |
dcode doctor | Check Node, your PATH, sign-in, the browser and git |
/model | Pick Raven Max or Raven Flash and the thinking level |
/crew | See what each agent of the latest crew did |
/review | Review local changes or a branch |
/rewind | Inspect or restore workspace checkpoints |
Models
DCode runs on Raven, the model family from Dipole ML Inc. Switch models and thinking levels any time from the model picker in the app, or with /model in the terminal. You can also add your own provider and API key in Settings.
Pick a model, then how hard it should think.
Raven Max · HighBest for large refactors, planning and bugs that hide across files.
Private by design
DCode is built in Toronto by Dipole ML Inc., a Canadian sovereign frontier AI lab. Privacy is part of the product, not a setting you have to find.
Prompts and diffs are stored encrypted at rest, and your code is never sold or shared.
Passwords are encrypted on your device and filled in by the app itself. They are never shown to the model or written to logs.
Posting, sending, buying and deleting on a website need your confirmation. So do pushing to git, deploying and touching credentials.
macOS builds are signed and notarized by Apple. Plugins are hash-verified on every download. DCode updates itself in the app.
Free to download. Sign in with your Dipole ML Inc. account, and a free account is enough to start.
curl -fsSL https://trydcode.app/install.sh | sh
Paste it in Terminal. It checks Node.js, installs DCode and opens it in the same window.
irm https://trydcode.app/install.ps1 | iex
Paste it in PowerShell. It checks Node.js, installs DCode and opens it in the same window.
npm install -g @dipoleml/dcode-cli
dcode
Works anywhere Node.js 22.13 or later runs. The second line starts DCode.
The CLI needs Node.js 22.13 or later. Not sure it is set up right? Run dcode doctor.
Run the installer, or paste the one-line command. Updates install themselves.
Use Google, Apple or email. One account covers the desktop app and the CLI.
Open a project, pick Raven Flash or Raven Max, and say what you need done.
| What it is | An agentic development environment (ADE): a desktop app and a CLI where AI agents plan, write, run and verify code |
|---|---|
| Made by | Dipole ML Inc., Toronto, Canada |
| Latest version | 2.5.0, released |
| Desktop app | Windows 10 or later (x64, ARM64) and macOS 13 or later (Apple silicon, Intel) |
| Terminal app | Windows, macOS and Linux with Node.js 22.13 or later |
| Models | Raven Max and Raven Flash, with thinking set to Off, Low, Standard or High |
| Parallel agents | Up to 24 in one crew |
| Price | Free download. A free account works. The DCode Coding Plan adds more usage for $29 USD per month |
Questions
Still deciding? The docs go deeper, and a person answers support.
DCode is an agentic development environment (ADE) made by Dipole ML Inc. You describe a task in plain language and DCode plans it, splits it across a crew of AI agents that work in parallel, edits your files, runs your build and tests, and tells you exactly what it verified. It runs as a desktop app on Windows and macOS and as a full-screen terminal app anywhere Node.js runs.
An agentic development environment is a tool where AI agents do the engineering work end to end instead of only suggesting code. In DCode the agents read your repository, plan the change, edit files, run shell commands, read the errors, fix what fails and verify the result with your own build and tests. You review and approve instead of typing every line.
Download the desktop installer for Windows or macOS from this page, or install the CLI with one command. On macOS and Linux run curl -fsSL https://trydcode.app/install.sh | sh. In Windows PowerShell run irm https://trydcode.app/install.ps1 | iex. Each command installs DCode and starts it. You can also run npm install -g @dipoleml/dcode-cli and then dcode.
Yes. DCode is free to download and a free Dipole ML Inc. account is enough to start. For heavy daily coding, the DCode Coding Plan adds a larger usage allowance for $29 USD per month. Current plans are listed on the pricing page.
A crew is a group of AI agents that DCode starts for a large task. A crew lead splits the task into independent parts, writes a shared brief so the parts fit together, and hands each part to an agent. Up to 24 agents can work in parallel. A reviewer then runs your real build and tests before the work is reported.
The desktop app runs on Windows 10 or later (x64 and ARM64) and on macOS 13 or later (Apple silicon and Intel). The terminal app runs on Windows, macOS and Linux with Node.js 22.13 or later.
DCode runs on Raven, the model family built by Dipole ML Inc. Raven Max is for hard engineering work, large refactors and planning. Raven Flash is for fast everyday edits and quick fixes. You choose how hard the model thinks: Off, Low, Standard or High. You can also add your own provider and API key in Settings.
Prompts and diffs are stored encrypted at rest, and your code is never sold or shared. Sensitive operations such as pushing to git, deploying or touching credentials ask you first unless you choose Full access. The details are in the Privacy Policy.
Yes. dcode -p "your task" runs one task and prints the result. Set DCODE_API_KEY to sign in without a browser, and use flags such as --max-turns, --json-schema, --allowed-tools and --mcp-config to control the run. dcode schedule runs a task on a schedule.
Yes. DCode supports the Model Context Protocol natively, so you can connect a server in one click and use its tools in the same conversation. The official marketplace has more than 330 plugins and skills, each hash-verified on every download, and you can write your own.
DCode is made by Dipole ML Inc., a Toronto-based Canadian sovereign frontier AI lab. Dipole ML Inc. also builds the Raven model family and the DipoleML API. DCode is proprietary software, and the desktop app and CLI are free to download.