DreamCoder CLI
DreamCoder’s own CLI runs coding tasks in a local workspace using the model provider selected by the user.
- 504 ★ on the repository
- Last commit
Broad access
Why this level
- The coding agent accesses files and the terminal; optional Computer Use controls applications and browsers.
Install
Manual install
bun installFirst dependency command in the source README after cloning. It covers only the monorepo root; complete the platform steps, including separate desktop dependencies.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The coding agent accesses files and the terminal; optional Computer Use controls applications and browsers.
This is third-party code. Review the repository files before installing.
What it does
DreamCoder CLI is part of the local DreamCoder workspace and starts a coding agent with the selected model. The agent works with files, terminal commands and connected tools while users inspect sessions and changes. Provider configuration is separate from the project. The repository includes its own dreamcoder entry point and session service. Desktop and browser interfaces expose the work; the documented browser workflow runs without building the desktop app.
Who it is for. Prototype builders who want to choose their model provider and inspect coding-agent activity in a local workspace.
Good fit when
- A test project needs comparison across configured model providers.
- A web-development result should be visible beside agent chat.
- You want to test the documented Ubuntu browser workflow.
- A long session needs visible file changes and tool activity.
Not a fit when
- You require an official packaged Linux desktop installer.
- The agent must not access project files and commands.
- You assume a local interface keeps cloud-model requests on the computer.
Example request
Prepare DreamCoder for a test project with local Ollama and show the actual working folder before edits.Limitations
Source setup needs Bun and Node.js, with separate root and desktop dependencies. Packaged desktop installers target macOS Apple Silicon and Windows x64; Linux uses a separately documented browser workflow. Cloud providers receive submitted context, while keys are stored in local files, so protect the configuration directory. A session’s working folder can differ from the server startup directory; verify the project path before assigning work.
How to disable. Stop the task in the interface, then stop the server and client from their terminals or app. Closing the browser tab alone does not stop session services.
FAQ
Where do I choose a model?
Add a provider under Settings, Providers and select the default model.
Why is the page open but the server unavailable?
Browser mode needs both backend and frontend processes. Check the frontend’s configured backend address.
Where is configuration stored?
Normally in `~/.dreamcoder/`; `DREAMCODER_CONFIG_DIR` can select another directory.
Can it control the desktop?
Yes, through the separate Computer Use feature. It is verified on macOS; Windows still needs community testing, and Linux support is a preview.
How do I connect a phone at home?
Enable H5 Access and use its QR code on the same LAN while keeping the desktop app running.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI runs a coding agent with files, shell commands and memory, while YAML configuration defines the model and connected tools.
A CLI agent that works with project files and shell commands, lets you choose a model provider and connects external MCP tools.
A terminal agent explores projects, prepares plans and makes edits or runs commands under a chosen approval policy.
OpenAI's coding agent that runs locally in your terminal, with IDE and desktop versions
A CLI connects installed coding agents into a team, dispatches work between sessions and collects results across machines.
A coding agent tuned for small local models, with a minimal system prompt, a planning mode and background sub-agents for research
README in short
Users can configure MCP servers visually. Design artifacts stay in the project directory for later editing.