serve

LlamaAgents

llamactl Reference

Serve your app locally for development and testing. Reads configuration from your project (e.g., pyproject.toml or llama_agents.yaml) and starts the Python API server, optionally proxying your UI in dev.

Usage

llamactl serve [DEPLOYMENT_FILE] [options]

DEPLOYMENT_FILE defaults to . (current directory). Provide a path to a specific deployment file or directory if needed.

Options

--no-install: Skip installing Python and JS dependencies
--no-reload: Disable API server auto‑reload on code changes
--no-open-browser: Do not open the browser automatically
--preview: Build the UI to static files and serve them (production‑like)
--port <int>: Port for the API server
--ui-port <int>: Port for the UI proxy in dev
--log-level <DEBUG|INFO|WARNING|ERROR|CRITICAL>: Log level for the API server
--log-format <console|json>: Log format for the API server
--persistence <memory|local|cloud>: Persistence mode for the workflow server. Defaults to local persistence.
--local-persistence-path <path>: SQLite database path for local persistence
--host <host>: Host for the API server. Defaults to 127.0.0.1; use 0.0.0.0 to accept remote connections.

Behavior

Prepares the server environment (installs dependencies unless --no-install)
In dev mode (default), proxies your UI dev server and reloads on change
In preview mode, builds the UI to static files and serves them without a proxy
Uses local workflow persistence by default; --persistence cloud stores workflow state in LlamaCloud

Credential injection

If your app uses LlamaCloud (e.g., for LlamaParse or cloud persistence), llamactl serve automatically injects credentials into the child process environment. It checks for LLAMA_CLOUD_API_KEY in the environment first, then falls back to the active profile’s API key. Credentials are also forwarded with PUBLIC_, VITE_, and NEXT_PUBLIC_ prefixes so frontend frameworks can access them during local development.

See llamactl auth for details on environment variable and profile-based authentication.