Skip to content

Getting Started

Requirements

  • Go 1.26+ if building from source
  • An API key for at least one supported runtime provider: Anthropic, OpenAI, or Gemini

Installation

Option 1 - Download Binary

Download the pre-built binary for your platform from the releases page.

Platform File
Linux (amd64) bolt-cowork-linux-amd64
macOS (amd64) bolt-cowork-darwin-amd64
macOS (arm64) bolt-cowork-darwin-arm64
Windows (amd64) bolt-cowork-windows-amd64.exe

Option 2 - Build from Source

git clone https://github.com/halukerenozlu/bolt-cowork.git
cd bolt-cowork
make build
# Binary is created at dist/bolt-cowork

Option 3 - Install via Go

go install github.com/halukerenozlu/bolt-cowork/cmd/bolt-cowork@latest

First Run

bolt-cowork --dir ./my-project "List all Go files"

If ~/.bolt-cowork/config.yaml does not exist, Bolt Cowork starts the TUI setup wizard. The wizard asks you to choose a provider and enter an API key. The key is stored in the system keyring, not in the config file.

After setup, Bolt Cowork will:

  1. Ask whether you trust the selected workspace directory
  2. Match relevant skills when needed
  3. Show you a plan
  4. Wait for your approval or revision
  5. Execute the task step by step
  6. Report the result

Runtime Providers

Bolt Cowork currently supports:

Provider Example models
Anthropic claude-opus-4-6, claude-sonnet-4-6
OpenAI gpt-4o, gpt-4o-mini, gpt-4.1
Gemini gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash

You can switch models during a session with /model. Bolt Cowork auto-detects the provider from known model names.


Approval Modes

Flag Behavior
--approval full Prompt for skill matching, plan approval, and steps
--approval plan-only Prompt at the plan stage only
--approval dangerous-only Prompt for dangerous execution steps
--approval none Run without prompts

The interactive /mode build shortcut maps to dangerous-only.


Common Commands

Command Purpose
/help Show available commands
/model Switch model and provider
/key Manage provider API keys
/config Inspect or adjust configuration
/dir Change workspace directory
/clear Clear conversation history and counters
/skills Show loaded skills
/skill <name> Inspect one skill
/use <name> Activate a skill for the next task
/mode Change approval mode
/init Initialize project context

In the TUI, press Ctrl+P for the same commands as a command palette, or type / directly into the chat input to get a live, filterable suggestion dropdown (↑/↓ to navigate, Tab to complete, Esc to dismiss without losing what you typed). The palette also exposes connect-provider (open the connection wizard for any provider), replace-credential (replace a stored API key after re-verifying it), and remove-credential (delete a stored API key) — available on both the welcome screen and inside a session.


Version History

Version Summary Status
v0.1 Core agent: sandbox, LLM provider, fallback chain, file operations Done
v0.2 Skill system: SKILL.md loading, keyword matching, prompt injection Done
v0.3 MCP client: JSON-RPC 2.0, tool discovery, permission profiles, resources, notifications Done
v0.4.0 TUI foundation with Bubble Tea, welcome screen, and split layout Done
v0.4.1 Agent integration, streaming output, plan widget, execution log, and command palette Done
v0.4.2 Palette overlay, grouped commands, ctrl+x chords, git status, and live right panel Done
v0.4.3 Modal system, setup wizard, keyring, animations, approval modal, and trusted directories Done
v0.4.4 Provider correctness, 4-step connection wizard, OpenRouter/DeepSeek/Mistral/Groq/Ollama/LM Studio support, SSE streaming, persistent sessions, native PDF merge/split Done
v0.4.5 TUI feedback fixes: no duplicate plan/result output for single-step runs, welcome-screen connection wizard, credential replace/remove, slash-command suggestions, one-entry-per-line list results Done