Quickstart
This guide takes you from a clean macOS or Linux machine to a working Anyy setup, a first conversation in Dashboard, and proof that the conversation survives a restart of the Dashboard process.
Who this is for
Use this guide if you:
- want to run Anyy for yourself on macOS or Linux;
- prefer Dashboard for everyday use, but still want the terminal commands behind it;
- have a provider API key, a supported subscription login, or a local Ollama or LM Studio server; and
- want one reliable chat working before adding messaging, automation, Skills, or external tools.
The fastest path
| Goal | First action | Next action |
|---|---|---|
| Use the full visual interface | Install Anyy | Run anyy setup, then anyy dashboard |
| Chat in a terminal | Install and configure Anyy | Run anyy |
| Send one prompt from a script | Start the resident Gateway | Run anyy chat "..." |
| Connect Telegram or another messaging platform | Verify a local chat first | Run anyy setup channel |
Configure one provider and complete one plain conversation before enabling fallback routing, messaging, Cron, MCP servers, or Browser. This keeps first-run problems small and easy to diagnose.
1. Install Anyy
macOS or Linux
Run the canonical native installer:
curl -fsSL https://anyy.ai/install.sh | sh
Then verify that your shell resolves the installed command:
anyy --version
The installer downloads the native release, verifies it against CHECKSUMS.sha256, activates it under the Anyy home directory, and creates a command entrypoint when it can do so safely.
The release installer supports macOS and Linux on amd64/x86_64 and arm64/aarch64. It does not start the Gateway or create your Profile data; setup does that only when needed.
If anyy is not on PATH, use the Binary: path printed by the installer to verify the installation, then add the exact directory shown in the installer's PATH guidance. For a normal non-root install, the command entrypoint is placed in $HOME/.local/bin when possible.
2. Choose a Provider
Start the interactive setup wizard:
anyy setup
For the shortest route to a first chat:
- Choose Quick setup.
- Select a provider and its authentication method.
- Select a model from the discovered catalog, or enter a model when the provider requires manual selection.
- Answer No to Set up model fallback routing? for now.
- Choose Skip when asked whether to connect a messaging platform.
- Review the provider, model, endpoint, and secret count. Confirm Write these changes?.
Setup supports several kinds of connection:
| What you already have | What to choose in setup |
|---|---|
| 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 Ollama or LM Studio | Ollama or LM Studio; start the local model server first |
| Another compatible service | OpenAI-compatible or Anthropic-compatible, then supply its endpoint |
The wizard also lists other currently usable providers. Its provider catalog is the authoritative list for the installed Anyy version.
When setup succeeds, it prints setup complete. Check the result with:
anyy config
This command shows the active Profile, Home, configuration file, database, provider, model, endpoint, and whether each credential reference resolves. It does not print secret values.
How settings are stored
Anyy keeps each Profile self-contained:
- provider and model settings are written to
<PROFILE_HOME>/config.yaml; - Anyy-managed credentials are stored separately under
<PROFILE_HOME>/secrets/with file mode0600; and - sessions and runtime state are stored in
<PROFILE_HOME>/state.db.
For the default non-root installation, <PROFILE_HOME> is normally $HOME/.anyy. When installed as root, it is /root/.anyy. Run anyy config instead of guessing when ANYY_HOME or named Profiles are involved.
The wizard stages its changes and shows a final review. Canceling before Write these changes? leaves the staged configuration unwritten.
3. Run Your First Chat
Dashboard is the recommended first surface because it puts chat, sessions, providers, tools, approvals, memory, Roles, Profiles, Channels, Skills, MCP, and Cron in one interface.
Start it from an interactive terminal:
anyy dashboard
If the resident Gateway is not running, Anyy shows Gateway required and asks:
Start Anyy Gateway now? [Y/n]:
Press Enter to accept. Anyy installs and starts the background service if necessary, waits for it to become ready, and then continues opening Dashboard. The Gateway remains available after the Dashboard process exits.
By default, Anyy prints a This machine URL and any usable Local network URLs on port 8767. An interactive run opens the This machine URL automatically. Use --no-open if you only want the URLs printed:
anyy dashboard --no-open
To bind only one explicit loopback address instead of automatic local-network listeners, run:
anyy dashboard --listen 127.0.0.1:8767
The printed URLs contain temporary access tokens. Do not paste them into chat, logs, issues, or screenshots. Press Ctrl+C in the terminal to stop the Dashboard listener.
Send the first message
In Dashboard:
- Open Chat.
- Select New session.
- Send this prompt:
Introduce yourself as Anyy in one sentence, then list three things you can help me do.
You have a working setup when the answer appears and the session is listed in the session panel.
The request can consume provider quota or incur provider charges. Do not include API keys, access tokens, or other secrets in a prompt.
4. Verify Sessions Work
Do not stop at a single response. Verify that the conversation is persistent:
- Send a follow-up message in the same session.
- Press
Ctrl+Cin the terminal that is running Dashboard. - Run
anyy dashboardagain. - Open Chat and select the earlier session.
- Confirm that both turns are present, then send one more follow-up.
This checks three separate paths: session storage, Dashboard reconnection, and continued execution through the resident Gateway.
For the terminal equivalent, run anyy, enter /sessions, and choose a previous conversation. If you already know the session ID, you can open it directly with:
anyy --resume SESSION_ID
5. Try Key Features
Dashboard controls
Use the navigation to inspect the setup you just created:
- Overview shows runtime health and the selected Profile.
- Models shows provider connections, authentication state, available models, and the active model.
- Chat keeps normal conversations separate from background sessions.
- Approvals shows work waiting for a decision and prior decisions.
- Roles and Memory control identity and durable personal context.
- Tools, Skills, MCP, Channel, and Cron are where you add capabilities after the base chat is stable.
Inside Chat, use Shift+Enter for a newline and Stop to interrupt an active response. Type / to open the available command menu.
Terminal interface
Open the TUI with:
anyy
Useful commands include:
/help Show commands and keyboard shortcuts
/new Start a new session
/sessions Open recent sessions
/status Show the current conversation state
/model Inspect or change the active model
/provider Inspect providers and provider authentication
/usage Show provider usage and context
/skills Insert an available Skill command
Press Enter to submit, Shift+Enter or Alt+Enter to insert a newline, Ctrl+C to interrupt an active turn or clear a draft, and Ctrl+D to exit.
One-shot CLI
With the resident Gateway running, send one prompt without opening a full interface:
anyy chat "Give me a five-line summary of what Anyy can do."
Add --json for machine-readable output. State-changing work that opens an approval is denied by default in non-interactive chat; --yes approves supported pending approvals, so use it only when you have reviewed the task and trust its effects.
6. Add the Next Layer
Add only the capability you need next:
| Capability | Start here | What it adds |
|---|---|---|
| Messaging | anyy setup channel | Connect a supported messaging platform and configure its access policy |
| Model fallback | anyy setup model fallback | Add an ordered backup provider/model chain |
| Web, image, video, or voice backends | anyy tools | Configure optional capability services |
| Browser | anyy browser connect | Connect an isolated local browser after explicit confirmation |
| Skills | anyy skills list | Inspect installed Skills before adding more |
| MCP | anyy mcp --help | Add, authenticate, probe, enable, and call MCP servers and tools |
| Scheduled work | anyy cron --help | Create and manage recurring jobs with approval or yolo execution mode |
Dashboard is an on-demand web sidecar. Gateway is the resident runtime. Stopping Dashboard with Ctrl+C closes the web listener; it does not stop a Gateway installed as a background service.
Common Failure Modes
| Symptom | Likely cause | Fix |
|---|---|---|
anyy: command not found | The command entrypoint directory is not on PATH | Follow the installer output's exact PATH instruction, reopen the shell, and run anyy --version |
| Setup exits after an empty API-key prompt | Empty input cancels first-time API-key setup | Run anyy setup model again and enter a valid key, or choose a subscription/local provider |
| Ollama or LM Studio models are unavailable | The local model server is not running or the endpoint is wrong | Start the local server, then rerun anyy setup model |
| Dashboard prints Gateway required | No resident Gateway is reachable | Accept the prompt, or run anyy gateway start and then anyy dashboard |
127.0.0.1:8767 is already in use | Another Dashboard or process owns the port | Run anyy dashboard --status, stop an old Dashboard with anyy dashboard --stop, or use --listen 127.0.0.1:8768 |
| Dashboard opens but chat cannot send | Provider, model, authentication, or Gateway health is invalid | Run anyy config, anyy status, and anyy doctor; then rerun anyy setup model if needed |
anyy chat cannot connect | One-shot chat does not start the resident Gateway | Run anyy gateway start first |
| A saved session seems missing | A different Profile or Home is selected | Compare the Profile and Home shown by Dashboard with anyy config |
Recovery Toolkit
Use this order so that each command narrows the problem before you change anything:
-
Check the selected runtime and Profile:
anyy statusanyy config -
Run the built-in diagnostics:
anyy doctor -
If the Gateway is unhealthy and no turn is active, restart it:
anyy gateway restart -
If provider authentication or model selection is wrong, repair only that layer:
anyy setup model -
If the Dashboard listener is the problem, inspect and stop old Dashboard processes, then start a fresh one:
anyy dashboard --statusanyy dashboard --stopanyy dashboard -
If multiple Profiles exist, confirm which one is active:
anyy profile list
Quick Reference
| Command | Purpose |
|---|---|
anyy | Open the interactive terminal interface |
anyy dashboard | Start Dashboard and open the This machine URL |
anyy dashboard --no-open | Start Dashboard without opening a browser |
anyy chat "PROMPT" | Send one prompt through the resident Gateway |
anyy --resume SESSION_ID | Open an existing session in the TUI |
anyy status | Show runtime and Gateway status |
anyy doctor | Run configuration and runtime diagnostics |
anyy config | Show resolved paths, provider, model, and credential status |
anyy gateway start | Install and start the resident Gateway service |
anyy gateway restart | Restart the resident Gateway service |
anyy setup model | Configure provider, authentication, model, and optional fallback routing |
anyy setup channel | Configure messaging platforms |
anyy update --check | Check whether a newer native release is available |
Next Steps
- Learn the visual control surface in Dashboard.
- Add or change a provider in Configuring models.
- Learn terminal commands and keyboard controls in TUI.
- Connect a messaging platform in Messaging.
- Survey the rest of the product in Features.
- Look up exact flags in the CLI command reference.