Your terminal, now with a full AI engineering team inside it.

💬 Questions, ideas, or just want to say hi? Join the Discussions

---

Why Bolt?

Bolt reads your codebase, understands what you're building, and ships code with you: in the terminal, and as a desktop app.

• Terminal-first. A fast, keyboard-driven TUI built with Effect, OpenTUI, and SolidJS.
• A whole team of agents. Ten specialized agents (code, plan, ask, review, debug, refactor, docs, security, migrate, perf) plus subagents for research and multi-step tasks.
• Sessions you can leave and re-join. Detach, attach (even to a remote server), fork, share, export, and import your work at any time.
• Extensible by design. MCP servers with OAuth, plugins, GitHub integration, and a built-in API server and web UI.
• Desktop app. The same engine wrapped in a multi-window desktop experience with tabs, an integrated terminal, and auto-updates.

Get started in 30 seconds

# Quick install (macOS / Linux)
curl -fsSL https://raw.githubusercontent.com/bolt-builder/bolt-cli/dev/install | bash

# Or with npm / bun / pnpm / yarn
npm i -g @bolt-builder/bolt-cli

Then, from any project:

# Open the TUI in the current directory
bolt

# Run a prompt directly (non-interactive)
bolt run "explain this codebase"

# Attach to a running server (local or remote)
bolt attach <url>

# Start a session with a specific agent
bolt run --agent ask "what does this project do?"

That's it. Bolt picks up your project context automatically.

Agents

Built-in agents. Switch with Tab in the TUI.

Agent	Access	Description
code	Full	Default: reads, writes, runs code
plan	Read	Explores and produces a plan without changing code
ask	Read	Questions and research, no file edits
code-review	Read	Reviews changes for correctness, style, and security
debug	Full	Debugs failing tests, crashes, and logic errors
refactor	Full	Safe refactoring with test verification at each step
docs	Edit	Writes and updates documentation and comments
security	Read	Security audit: vulnerabilities, secrets, patterns
migrate	Full	Framework upgrades and dependency migrations
perf	Full	Performance analysis and optimization

Subagents for delegation: @general for complex multi-step tasks and @explore for fast read-only codebase research. You can also define your own agents as markdown files or generate one with bolt agent create.

Power features

• Remote sessions: bolt attach <url> and bolt run --attach <url> work against any running server, with basic auth and mDNS discovery (serve --mdns).
• Fork and share: branch a session with --fork, share with run --share, import straight from a share URL with bolt import <url>, export sanitized transcripts with export --sanitize.
• Model control: pick reasoning effort with run --variant, show reasoning with --thinking, stream raw events with --format json.
• MCP with OAuth: bolt mcp add/auth/logout/debug manages servers end to end, including OAuth flows and connectivity probes.
• Minimal mode: --mini gives a compact interactive UI on bolt, run, and attach.
• Local data, inspectable: your sessions live in sqlite; query them with bolt db any time.

Desktop app

Bolt also ships as a desktop app: multi-window and multi-tab session management, a prompt composer with attachments and clipboard image paste, an integrated terminal and file tree, a command palette, full settings (keybinds, models, providers, servers), native notifications, deep links, WSL integration on Windows, auto-updates across dev/beta/prod channels, and 17 UI languages. Download it from the releases page.

Tip

On macOS you can install it in one line, no Gatekeeper detour:

curl -fsSL https://raw.githubusercontent.com/bolt-builder/bolt-cli/dev/install-desktop | bash

Configuration

Configuration lives in .bolt/bolt.jsonc in your project root (created on first run).

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    // Provider config goes here
  },
}

Environment variables

Variable	Description
BOLT_LOG_LEVEL	Log level: DEBUG, INFO, WARN, ERROR
BOLT_PRINT_LOGS	Print logs to stderr
BOLT_PURE	Run without external plugins
BOLT_SERVER_PASSWORD	Basic auth password for the server
BOLT_SERVER_USERNAME	Basic auth username for the server

All CLI commands

Command	Description
bolt	Launch the interactive TUI
run	Run a non-interactive prompt
attach	Attach the TUI to a running server
session	Manage sessions (list, delete)
agent	List and create agents
providers	Manage LLM provider credentials (alias: auth)
models	List available models
mcp	Manage MCP servers (add, list, auth, logout, debug)
serve	Start the headless API server
web	Start the server and open the web UI
upgrade	Upgrade to the latest version
uninstall	Remove bolt
completion	Generate shell completions
export	Export session history (with optional --sanitize)
import	Import a session from a file or share URL
plugin	Install and manage plugins
github	GitHub Actions agent (install, run)
pr	Check out a GitHub PR into a local branch
commit	Commit staged changes with a generated message
review	AI review of a diff with pass/fail exit codes
stats	Show token usage and cost statistics
db	Query the local session database
debug	Troubleshooting tools (config, lsp, snapshots, and more)
acp	Agent Client Protocol server for editors (e.g. Zed)

Roadmap

The roadmap is about one thing now: making the agent itself smarter, more autonomous, and more fun to work with. Have an opinion? Tell us in Discussions.

Now

• Multi-agent pipelines: one agent plans, one codes, one reviews; a tiny eng team in your terminal
• Automatic agent selection: Bolt reads your prompt and quietly routes it to the right specialist

Under the hood, the session sync engine and V2 session core keep rolling out. They are the plumbing that makes everything below possible.

Next

Agents that remember (4)

• Repo convention learning: the agent picks up your codebase's style and sticks to it
• Compounding memory: every session teaches the next one
• Memory introspection: ask the agent why it believes something and where it learned it
• Cross-session recall: "do it like we did in that auth refactor last month"

Agents that check their work (3)

• Test-aware refactor loops: change, run, verify, repeat until green
• Confidence scoring: the agent tells you when it is guessing
• Self-review pass before any diff is handed to you

Agents that never sleep (5)

• Watch mode: tests rerun on every save and failures fix themselves
• bolt cron: scheduled agent chores (dependency bumps, changelog drafts, issue triage)
• On-red-main automation: bisect, blame, and propose the fix before you've seen the alert
• Flaky test detection and quarantining
• Background job queue: run --background plus bolt jobs list/tail/kill

Agents with guardrails (4)

• Guardrail agent that vetoes risky commands before they run
• Budget guards: --max-cost and --max-tokens on any run
• bolt undo: one-command rollback when an experiment goes sideways
• Named checkpoints: save points you can rewind the repo and the conversation to

Party tricks (4)

• Design-to-code: hand the agent a Figma file, get components back
• Codebase visualization maps drawn by the agent
• Pair-programming mode with a shared cursor
• Session replays you can share like a highlight reel

Later

• Autonomous long-horizon projects spanning days
• Fine-tuned repo-specific models
• On-device small-model routing for trivial tasks
• Agent federation across organizations
• AI release manager: cut, verify, and publish releases end to end

Done

• bolt review --staged / --branch: one-shot AI diff review with exit codes
• bolt commit: commit messages you don't have to rewrite
• Best-of-N runs: fire the same task at several models in parallel, rank the results, keep the winner
• Push-to-talk voice input in the TUI
• Agent evaluation benchmark harness (bolt eval)
• Project memory module: automatic capture with save and recall tools
• bolt memory: view and edit what the agent remembers per project (/memory in the TUI)
• Per-session memory controls
• /plan mode, /todos task list, and /btw side questions that don't interrupt the task
• bolt logs with tail and follow
• .boltignore-aware agent file search
• macOS desktop app with a one-line terminal installer (no Gatekeeper detour)

Contributing

We'd love your help, whether it's a bug fix, a new provider, or better docs. Read the contributing guide to get started, or pick up a good first issue.

# Clone and build
git clone https://github.com/Bolt-builder/bolt-cli.git
cd bolt-cli
bun install

# Start the TUI in dev mode (from packages/bolt)
cd packages/bolt
bun dev

# Run tests (from package directories)
bun test

# Type-check
bun typecheck

Enjoying Bolt?

Consider giving it a star ⭐ — it helps other developers discover the project, and it's the easiest way to support the work if you can't contribute code right now.

License

PolyForm Noncommercial 1.0.0 License © Bolt CLI