Skip to content
API Blog

Owletto CLI Reference

The owletto CLI starts local Owletto runtimes, installs bundled Owletto starter skills, configures agent clients, and runs Owletto tools directly.

Terminal window
# Run without installing
npx owletto@latest <command>
# Or install globally
npm install -g owletto
owletto <command>

The published package name is owletto.

Starts a local Owletto runtime.

Terminal window
npx owletto@latest start
npx owletto@latest start --port 8787

Default behavior:

  • listens on http://localhost:8787
  • uses embedded Postgres (PGlite) by default
  • stores local data in ~/.owletto/data/ for packaged installs
  • uses ./data/ when running from an owletto repo checkout

If DATABASE_URL is set, the CLI starts the server against external Postgres instead.

Installs bundled Owletto starter skills into a local skills/ directory.

Terminal window
npx owletto@latest skills list
npx owletto@latest skills add owletto
npx owletto@latest skills add owletto-openclaw

Use:

  • owletto for generic Owletto memory and tool workflows
  • owletto-openclaw for OpenClaw-specific memory plugin setup

Configures local agent clients to use an Owletto MCP endpoint.

Terminal window
npx owletto@latest init
npx owletto@latest init --url http://localhost:8787/mcp

Detects supported clients and auto-configures them when possible. Falls back to manual steps when needed.

Authenticates the CLI against an Owletto MCP server using OAuth.

Terminal window
npx owletto@latest login https://app.lobu.ai/mcp

By default, the CLI opens a browser and completes an authorization-code flow with a local callback server.

Useful flags:

  • --device uses device-code login for headless environments or browserless agents
  • --noOpen prints the login URL instead of opening a browser
  • --scope overrides the requested OAuth scopes

Example for a headless box:

Terminal window
npx owletto@latest login https://app.lobu.ai/mcp --device

Prints a usable access token from the saved session.

Terminal window
npx owletto@latest token
npx owletto@latest token --raw

This is mainly useful for integrations or plugin setups that need a token command.

Checks that the saved session is valid and that the CLI can reach the MCP endpoint.

Terminal window
npx owletto@latest health

Owletto sessions are organization-aware. After login, set the default org if needed:

Terminal window
npx owletto@latest org current
npx owletto@latest org set my-org

You can also override organization and server selection per command:

  • --org <slug>
  • --url <mcp-url>
  • OWLETTO_ORG
  • OWLETTO_URL

Lists tools when called without arguments, or executes a tool when given a tool name and JSON params.

Terminal window
# List available tools
npx owletto@latest run
# Search knowledge
npx owletto@latest run search_knowledge '{"query":"Acme"}'
# Read saved content
npx owletto@latest run read_knowledge '{"query":"customer preferences"}'
# Save new knowledge
npx owletto@latest run save_knowledge '{"content":"Prefers weekly summaries","semantic_type":"preference","metadata":{}}'

This is the most direct way to inspect or test Owletto behavior outside an agent runtime.

The exact tool list depends on the endpoint and your session scope. Run owletto run with no arguments to see what is available.

Core memory: search_knowledge, read_knowledge, save_knowledge

Watchers: list_watchers, get_watcher

Organization: list_organizations, switch_organization (unscoped endpoint only)

Admin / workspace (admin sessions only): manage_entity, manage_entity_schema, manage_connections, manage_feeds, manage_auth_profiles, manage_operations, manage_watchers, manage_classifiers, query_sql

Checks local prerequisites such as Node, Docker, and current server reachability.

Terminal window
npx owletto@latest doctor

Captures browser-based auth or cookie state for connectors that rely on a real browser session.

This is mainly for connector setup, not day-to-day memory usage.

Writes OpenClaw plugin config for @lobu/owletto-openclaw using an owletto token command.

For most users, the shortest path is:

Terminal window
npx owletto@latest skills add owletto
npx owletto@latest init

That gives the agent both:

  • the Owletto skill so it knows how to use Owletto well
  • the MCP configuration so it can actually connect to Owletto

When working inside the owletto repository itself, you can run the TypeScript entrypoint directly:

Terminal window
pnpm -C packages/owletto-cli exec tsx src/bin.ts start
pnpm -C packages/owletto-cli exec tsx src/bin.ts skills list
pnpm -C packages/owletto-cli exec tsx src/bin.ts init
pnpm -C packages/owletto-cli exec tsx src/bin.ts run search_knowledge '{"query":"spotify"}'

Use the Lobu CLI to scaffold and run Lobu projects. Use the owletto CLI to install the Owletto skill, configure clients, and operate Owletto itself.