Setup

anyy setup is the guided way to turn an installed Anyy command into a working personal assistant. It can configure a model, connect messaging, add optional media and web capabilities, and register MCP servers.

Start with Quick setup

Choose Quick setup for a first installation. Get one local conversation working before adding fallback models, messaging, or external tools.

Before You Start

You need an interactive terminal and at least one usable model connection:

• a provider API key;
• a supported subscription login;
• a running Ollama or LM Studio server; or
• an OpenAI-compatible or Anthropic-compatible endpoint.

Run the wizard from the Profile you want to configure:

anyy setup

Use --profile NAME before the command, or --home PATH after it, when you do not want the default Profile:

anyy --profile work setup
anyy setup --home /path/to/anyy-home

Choose a Setup Mode

The first screen offers two paths:

Mode	Configures
Quick setup	Provider, authentication, main model, optional model fallback, and messaging
Full setup	Everything in Quick setup, followed by optional image, video, voice, web-search, and MCP configuration

Both paths finish with a single review. The collected configuration is committed after you confirm Write these changes?. An explicitly started sign-in can contact the authentication service before that confirmation.

Configure the Main Model

The Provider & Model section asks for:

1. a provider;
2. an authentication method offered for that provider;
3. any required endpoint or credential; and
4. the main model.

When the provider exposes a model catalog, Anyy fetches it and lets you select a model. Compatible or local providers can also require a manually entered model ID.

Common paths include:

Connection	Setup choice
OpenAI API key	OpenAI → OpenAI API key
ChatGPT subscription	OpenAI → ChatGPT subscription (Codex)
Anthropic API key	Anthropic → Anthropic API
Claude Code subscription	Anthropic → Claude Code subscription
Local model	Ollama or LM Studio
Compatible service	OpenAI-compatible or Anthropic-compatible

The installed wizard is the authoritative provider catalog. It only shows authentication methods implemented by that Anyy version.

Treat custom endpoints as trusted infrastructure

A compatible endpoint receives your prompts, attachments, and model request metadata. Use an endpoint you operate or trust.

Configure only the model

To change provider authentication or the main model without walking through messaging and optional tools, run:

anyy setup model

If a Gateway is already reachable, Anyy reloads provider configuration after the confirmed write. Otherwise, the saved configuration takes effect the next time the runtime starts.

Add Model Fallbacks

During full or quick setup, Anyy asks whether to configure fallback routing. Skip this on a first run unless you already have a second tested provider/model pair.

You can manage the ordered fallback chain later:

anyy setup model fallback

A main model must already be configured. Fallbacks are tried in their saved order when routing can move past an unavailable main target; they are not a substitute for valid credentials or a healthy provider.

Connect Messaging

The Messaging section lets you connect a supported platform or skip it. For a focused messaging-only run, use:

anyy setup channel

Channel setup may ask for account identifiers, access policies, webhook details, and credentials. If a channel change is confirmed, Anyy installs and starts the resident Gateway service so the connection can remain available after the terminal closes.

Webhook delivery also needs the Gateway webhook listener. It is disabled by default; enable channel_webhooks.enabled, choose a reachable channel_webhooks.listen address and restart the Gateway before configuring the platform callback. Saving or reloading an account alone does not enable this listener. See the platform guide for its authentication and public HTTPS requirements.

See Messaging Platforms for supported platforms, pairing, health checks, and reload commands.

Optional Capabilities

Full setup can also configure:

• image understanding;
• video understanding;
• image generation;
• text-to-speech and speech-to-text;
• web search; and
• one or more MCP servers.

Each item is optional. Select only capabilities you intend to use; each external backend can have its own credentials, data handling, and usage charges.

Verify media using the supported input surface: the Dashboard uploader classifies PNG/JPEG as images and other uploads as files. Merely uploading an audio or video file there does not invoke speech or video understanding. See Media Understanding.

Review and Write

Before writing, the wizard displays Review Changes. Check the provider, model, endpoint, credential count, fallback route, channels, optional capabilities, and MCP entries.

Confirm:

Write these changes? [Y/n]

On success, Anyy prints setup complete and suggests:

anyy status
anyy config
anyy

Canceling before final confirmation leaves collected local settings unwritten and does not submit the collected Channel actions. After confirmation, Channel actions execute directly through the resident Gateway; the wizard checks their local receipts and reports connection diagnostics separately.

Where Setup Stores Data

For the selected Profile home:

Data	Location
Provider, model, routing, and capability settings	<PROFILE_HOME>/config.yaml
Provider/MCP credential files	<PROFILE_HOME>/secrets/
Channel accounts and protected encrypted credentials	<PROFILE_HOME>/state.db
Sessions and runtime state	<PROFILE_HOME>/state.db

Credential files and encrypted Channel bindings are separate from config.yaml and are written with file mode 0600. Run anyy config to see resolved paths and credential status without printing secret values.

Focused Setup Commands

anyy setup [--home PATH]
anyy setup model [--home PATH]
anyy setup model fallback [--home PATH]
anyy setup channel [--home PATH]
anyy setup channel CHANNEL [--home PATH] \
  [--set INPUT=VALUE] \
  [--sensitive-fd INPUT=FD] \
  [--external-locator-fd INPUT=FD]

The last form is the non-interactive channel path. Sensitive values are read from file descriptors so they do not need to appear in command history or process arguments.

Troubleshooting

Problem	Fix
requires an interactive terminal	Run anyy setup, anyy setup model, or the interactive anyy setup channel directly in a terminal
Empty API-key input cancels setup	Run anyy setup model again and enter a valid key, or choose another authentication method
No local models appear	Start Ollama or LM Studio, verify its endpoint, and rerun model setup
Subscription login does not finish	Complete the browser/device flow, then rerun setup if the terminal session expired
Saved settings are not visible	Compare the active Profile and Home with anyy config; you may have configured another Profile
A channel does not connect	Run anyy channel doctor CHANNEL, then anyy channel reload after correcting its settings

Continue with Dashboard Overview or Configuring Models.