A workspace.
A model. A first step.
Start small: one project, a few source files, and a task you want to move forward.
Create a project
Give your work its own space. Import the files you want to work with; workspace tools are restricted to this project’s directory.
Choose how your model runs
Use Built-in Models in Settings to download a model for on-device inference. Or configure Ollama, LocalEngine, LM Studio, OpenAI, or an OpenAI-compatible provider. Select a model that fits your Mac and your task.
Add useful context
Import text or text-based PDFs into Knowledge. Start with keyword search, or configure an embedding model for semantic retrieval. Open citations to check the original source.
Ask, review, and continue
Start a conversation and follow the tool timeline. Inspect the diff before approving a file write. Use Tasks to run a workflow template and resume unfinished steps.
Bring your own tools, when you need them.
Add a Streamable HTTP MCP server in Settings, enable it for the project, and review each tool call before it runs. Remote MCP servers require HTTPS; stdio servers are not supported.
Explore all features