Meet Dash.
Your AI coding agent.

Code with AI

Dash runs an agent loop in your project folder: it reads files, plans the work, and applies edits while you review what happens. Ask it to fix a failing test, refactor a module, or explain a code path — tool rows show every read, edit, and search as it runs.

• Reads and edits files in your workspace
• Maintains a task checklist for multi-step work
• Rolls back edits via checkpoints

Choose Your AI Provider

Dash integrates the NVIDIA NIM API — an OpenAI-compatible endpoint at integrate.api.nvidia.com. The provider ships with fallback Nemotron models, and /models lists everything your API key can access. No provider account of ours — you bring your own key.

Start in Your Terminal

No IDE extension, no browser tab. Install once, cd into any project, and launch.

Developer-Controlled Workflow

Dash gives the model a toolset — 24 tools — but keeps the controls with you:

• Every shell command requires your explicit approval
• Reads, search and workspace edits run without ceremony
• File writes outside the workspace and destructive file operations always ask first
• Checkpoints record file state before edits, for reversible changes

Privacy and API Keys

Dash runs on your machine, and your project stays where it is. When you request AI features, prompts and the project context Dash assembles are sent to the AI provider you configured. Your API key is stored locally, encrypted (AES-256-GCM), at ~/.dash/config.json — never printed, never transmitted anywhere except as the bearer token to the provider endpoint.

Nvidia Support

Dash has first-class NVIDIA support. Connect an nvapi_ key via /providers, created from your NVIDIA account (build.nvidia.com). Dash talks to the NIM API, fetches the live model list for your account, and defaults to Nemotron models if the list cannot be fetched.