generated: '2026-08-27'
method: searched
source: https://localai.io/reference/cli-reference
docs: https://localai.io/reference/cli-reference
name: local-ai
description: >-
  LocalAI ships as a single Go binary named `local-ai` which is both the server and the CLI. Unusually,
  the CLI is not a client wrapper around a hosted API — `local-ai run` IS the API. Subcommands additionally
  expose inference, agents and the MCP admin server without a running HTTP server, and every flag has an
  equivalent environment variable (environment variables take precedence over flags).
install_methods:
- method: container
  recommended: true
  registry: docker.io/localai/localai (mirror quay.io/go-skynet/local-ai)
  docs: https://localai.io/installation/
- method: install script
  command: 'curl https://localai.io/install.sh | sh'
  probed_status: 200
- method: macos-dmg
  asset: LocalAI.dmg
  docs: https://localai.io/docs/getting-started/macos/
- method: linux-binary
  assets:
  - local-ai-v4.9.0-linux-amd64
  - local-ai-v4.9.0-linux-arm64
  - local-ai-v4.9.0-darwin-arm64
  docs: https://localai.io/installation/
- method: kubernetes
  chart: local-ai (https://go-skynet.github.io/helm-charts)
  docs: https://localai.io/installation/
  note: Chart 3.4.2 / appVersion 2.23, last published 2024-11-21 — see packages/localai-packages.yml.
- method: source
  command: 'GO_TAGS=auth make build'
  docs: https://localai.io/installation/
commands:
- name: run
  description: Start the LocalAI server and HTTP API.
- name: models
  description: Manage LocalAI models and definitions.
- name: backends
  description: Manage LocalAI backends and definitions.
- name: chat
  description: Run the built-in terminal agent against a LocalAI server.
- name: tts
  description: Convert text to speech.
- name: sound-generation
  description: Generate audio files from text or audio.
- name: transcript
  description: Convert audio to text.
- name: agent
  description: Run agents standalone without the full LocalAI server.
- name: mcp-server
  description: >-
    Run the LocalAI admin tool surface as a stdio MCP server controlling a remote LocalAI instance over
    HTTP. Flags --target <url>, --api-key <key>, --read-only. See mcp/localai-mcp.yml.
- name: worker
  description: Start a worker for distributed mode (generic, backend-agnostic).
- name: p2p-worker
  description: Run workers to distribute workload via p2p (llama.cpp only).
- name: agent-worker
  description: Start an agent worker for distributed mode (executes agent chats via NATS).
- name: util
  description: Utility commands.
- name: explorer
  description: Run the P2P explorer.
- name: federated
  description: Run LocalAI in federated mode.
- name: completion
  description: Generate shell completion scripts for bash, zsh or fish.
flag_groups:
- name: Global
  examples: ['--log-level (error|warn|info|debug|trace)', '--debug (DEPRECATED)']
- name: Storage
  examples: ['--models-path', '--data-path', '--generated-content-path', '--upload-path', '--localai-config-dir']
- name: Backend
- name: Models
- name: Performance
  examples: ['--f16']
- name: API
  examples: ['--address']
- name: Hardening
  examples: ['LOCALAI_OPAQUE_ERRORS', 'LOCALAI_DISABLE_API_KEY_REQUIREMENT_FOR_HTTP_GET']
- name: Authentication
  examples: ['LOCALAI_AUTH', 'LOCALAI_OIDC_ISSUER', '--disable-local-auth', 'LOCALAI_REGISTRATION_MODE']
- name: Chat
- name: P2P
  examples: ['--p2p', '--federated']
env_var_convention:
  prefix: LOCALAI_
  precedence: Environment variables take precedence over CLI flags.
  dotenv: Supported via .env files; see https://localai.io/advanced/
shell_completion:
  shells: [bash, zsh, fish]
  command: local-ai completion
  docs: https://localai.io/reference/shell-completion