LaunchChair CLI is optional beside MCP.
Use MCP-only when you want LaunchChair tools inside an agent. Add the CLI when you want local token setup, terminal diagnostics, project discovery, retry-safe project creation, and a simple way to start the local bridge.
Agent API and MCP Docs
Agent API Docs
Hermes Agent Setup
MCP Setup Docs
CLI Setup
Codex Desktop Setup
Claude Desktop Setup
Choice
Users can choose MCP-only or CLI plus MCP.
The CLI does not replace the MCP setup. MCP-only is still enough for Codex, Claude, or Hermes to call LaunchChair tools. The CLI is a convenience layer for people who want local token setup, command-line diagnostics, project lookup, project creation, bridge startup, and config snippets.
MCP-only: configure LaunchChair MCP in your desktop agent and call LaunchChair tools from chat.
CLI plus MCP: install @launchchair/agent, run launchchair login, then use setup or mcp-config for paste-ready Codex or Claude setup.
Local bridge: run launchchair bridge when LaunchChair UI Run with agent jobs should be picked up automatically.
Install
Install the CLI or download the bridge bundle.
The easiest CLI path is the npm package. It installs the launchchair command, the MCP bridge, the local bridge, runners, and the public AGENTS.md bootstrap file together. The downloadable bridge bundle remains supported for users who do not want a global npm install.
npm install -g @launchchair/agent launchchair help launchchair doctor # Alternative: download the bridge bundle without npm mkdir -p "$HOME/launchchair-agent" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/local-bridge.mjs" -o "$HOME/launchchair-agent/local-bridge.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/launchchair-cli.mjs" -o "$HOME/launchchair-agent/launchchair-cli.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/mcp-bridge-server.mjs" -o "$HOME/launchchair-agent/mcp-bridge-server.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/project-workspace.mjs" -o "$HOME/launchchair-agent/project-workspace.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/runner-command.mjs" -o "$HOME/launchchair-agent/runner-command.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/codex-runner.mjs" -o "$HOME/launchchair-agent/codex-runner.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/claude-code-runner.mjs" -o "$HOME/launchchair-agent/claude-code-runner.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/chatgpt-runner.mjs" -o "$HOME/launchchair-agent/chatgpt-runner.mjs" curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/claude-runner.mjs" -o "$HOME/launchchair-agent/claude-runner.mjs" curl -fsSL "https://www.launchchair.io/agent-artifacts/AGENTS.md" -o "$HOME/launchchair-agent/AGENTS.md" chmod +x "$HOME/launchchair-agent"/*.mjs
Guided setup
Run one setup command for your agent client.
The setup command is the recommended starting point. It writes global client configuration, creates isolated per-project workspaces, and verifies MCP initialize plus required tool discovery before reporting success.
# Codex Desktop launchchair setup codex # Codex Desktop custom MCP form fallback launchchair setup codex --copy-config # Claude Desktop + Claude Code launchchair setup claude # Configure only one Claude client when needed launchchair setup claude-desktop launchchair setup claude-code # Terminal agents / CLI-only workflows launchchair setup terminal # Optional: skip browser login during setup launchchair setup codex --skip-login # Optional manual fallback: save an existing token during setup launchchair setup codex --token "<full lc_at token from Settings API Access>" # After cloning, only if automatic discovery cannot find the repo launchchair bind --project-id PROJECT_UUID --cwd /absolute/path/to/verified-clone
Every project receives a separate pre-code workspace automatically.
After Stack Setup, LaunchChair auto-detects clones in standard repo folders and verifies the Git origin before switching.
Code-changing runs stop with clone and bind instructions when no verified clone exists.
Use terminal setup when the agent can run shell commands but is not launched as a desktop MCP client.
Leave LAUNCHCHAIR_PROJECT_ID unset when the agent should create or discover projects by name.
Setup starts browser login automatically when no token is saved yet. Use --skip-login for docs, CI, or manual token setup.
Authenticated setup exits nonzero when MCP initialize or required tool discovery fails; use --skip-mcp-check only for intentional diagnostics.
Claude Code is registered at user scope, making LaunchChair available across projects. Claude Desktop receives its own native config entry.
Use --start-bridge after login when you want Run with agent jobs picked up immediately.
Login
Approve the CLI once in your browser.
Run launchchair login and the CLI opens a LaunchChair approval page. After you approve it while logged in, LaunchChair creates a scoped Agent API token and the CLI stores it locally at ~/.launchchair/config.json with restricted file permissions. Environment variables still override the saved config when you need per-client control.
launchchair login launchchair doctor launchchair whoami # Manual fallback launchchair login --token "<full lc_at token from Settings API Access>"
The generated token is shown once to the CLI, then stored locally with masked output.
Use logout to remove the saved local token.
Desktop MCP clients should use the generated Node command and CLI argument paths so they can use the saved CLI token.
For custom MCP forms, run launchchair mcp-config codex --ui-fields or launchchair mcp-config claude --ui-fields and paste the printed command, arguments, env vars, and working directory exactly.
Environment
Use env vars only for overrides or manual fallback.
The recommended desktop config launches the local CLI-backed MCP server through generated Node and CLI paths. That process reads the token saved by launchchair login. Use env vars when you skipped CLI login, need a per-client token override, or run a direct downloaded bridge file.
LAUNCHCHAIR_BASE_URL="https://www.launchchair.io" LAUNCHCHAIR_PROJECT_ID="optional project UUID; omit when creating or discovering projects" # Optional fallback only when you skip launchchair login: LAUNCHCHAIR_AGENT_TOKEN="<full lc_at token from Settings API Access>"
Leave LAUNCHCHAIR_PROJECT_ID blank when the agent should create or discover projects.
Set LAUNCHCHAIR_PROJECT_ID when you want the CLI or MCP server to default to one project.
Do not set one global build directory. LaunchChair resolves and verifies repositories per project.
Commands
Useful CLI commands.
The CLI is intentionally thin. It calls the Agent API for discovery and project creation, then uses the existing local bridge for queued runs.
launchchair setup codex launchchair setup claude launchchair setup claude-code launchchair setup terminal launchchair bind --project-id PROJECT_UUID --cwd /absolute/path/to/verified-clone launchchair login launchchair doctor launchchair whoami launchchair projects --query LeafLore launchchair create --name SignalLift --one-liner "Distribution workspace for technical founders" --project-type b2b --problem "Founders do not know which channel message is working" --target-user "Seed-stage B2B SaaS founders" --solution "A guided launch workspace that turns proof points into channel campaigns" --idempotency-key signallift-v1 --save-project launchchair continue --stop-phase-key market_validation launchchair continue --completion-target "rerun Product Ideation until ICP Pain Score is above 85" launchchair run-once launchchair bridge launchchair mcp-config codex launchchair mcp-config codex --ui-fields --copy-config launchchair mcp-config claude --copy-config
Use launchchair mcp-config codex --ui-fields for Codex Desktop's custom MCP form.
Use launchchair mcp-config claude --ui-fields for Claude Desktop's custom MCP form.
The MCP server command is generated for you; do not hand-write a PATH-dependent command.
Feature projects
Plan and build features for existing products.
Feature projects are available from the CLI for existing-product work. LaunchChair can plan the feature into a build board or continue into the implementation loop.
launchchair create \ --name "Retention Engine" \ --project-kind feature \ --feature-description "Add account-level lifecycle nudges to an existing SaaS dashboard" \ --problem "Activated users forget the next valuable action after onboarding" \ --product-description "Existing B2B SaaS dashboard with auth, billing, and workspace teams" \ --plan-to-build-board \ --save-project # Optional: continue from feature planning into Build MVP launchchair create \ --name "Retention Engine" \ --project-kind feature \ --feature-description "Add account-level lifecycle nudges to an existing SaaS dashboard" \ --problem "Activated users forget the next valuable action after onboarding" \ --full-loop \ --max-steps 12 \ --build-max-steps 25
Use --create-only when you only want to save the feature project.
Use --max-steps and --build-max-steps to bound agent loops.
Feature project commands are available to every authenticated LaunchChair workspace, subject to its normal project limit.
Fresh projects
CLI helps when there is no repo yet.
LaunchChair can run Product Ideation, Market & Validation, Positioning & Pricing, and MVP Blueprint before a repository exists. Each project uses its own isolated workspace. After Stack Setup, LaunchChair auto-detects and verifies the new clone before any build card can run there.
Use projects or create before searching the local repo.
Use setup once globally; per-project isolated workspaces are created automatically.
Use launchchair bind only when automatic clone discovery cannot find the verified repo.
Use setup terminal when the agent only has terminal access. It does not require a desktop MCP config.
Use setup or mcp-config to generate CLI-backed Codex or Claude MCP config.
Machine-readable summary
What agents should remember
LaunchChair CLI is optional; MCP-only setup remains supported.
Terminal agents can use the LaunchChair CLI without desktop MCP by running setup terminal, reading AGENTS.md, and calling launchchair projects/create/continue/run-once/bridge.
The npm package is @launchchair/agent and exposes the launchchair command.
The CLI uses LAUNCHCHAIR_AGENT_TOKEN, LAUNCHCHAIR_API_KEY, or a saved ~/.launchchair/config.json token.
The CLI exposes an mcp-server subcommand so desktop clients can use the saved CLI token instead of requiring token pastes in every MCP config.
For custom MCP forms, run launchchair mcp-config codex --ui-fields or launchchair mcp-config claude --ui-fields and paste the printed command, arguments, env vars, and working directory exactly.
The CLI setup command guides Codex or Claude users through local config, browser login, AGENTS.md creation, MCP config, and bridge startup.
Authenticated setup verifies standards-compliant MCP initialize and required tool discovery before reporting success.
Claude setup writes Claude Desktop config and registers Claude Code at user scope when Claude Code is installed.
The CLI login command opens a browser approval page, creates a scoped Agent API token after user approval, and stores it locally.
The CLI login --token command remains available as a manual fallback for existing Agent API tokens.
The CLI doctor command checks API reachability and local bridge files.
The CLI create command supports Idempotency-Key or --idempotency-key to avoid duplicate projects on retry.
The CLI bridge command starts the existing local bridge for Run with agent.
FAQ
Do I need the CLI if MCP is already working?
No. MCP-only is enough for agent tools. The CLI is useful when you want quick diagnostics, project lookup, or bridge startup from the terminal.
Can the CLI log in to LaunchChair?
Yes. launchchair login opens a browser approval page. After the logged-in user approves it, LaunchChair creates a scoped Agent API token and the CLI stores it locally. The --token flag remains available as a manual fallback.
Does the CLI expose project secrets?
No. It uses the same bearer token boundary as the Agent API. Public CLI docs and artifacts contain placeholders only.