> ## Documentation Index > Fetch the complete documentation index at: https://docs.aion.to/llms.txt > Use this file to discover all available pages before exploring further. # Aion Chat > Standalone terminal chat for A2A-compatible agents. Aion Chat terminal interface Aion Chat is a terminal interface for chatting with A2A-compatible agents. It can run through the Python SDK CLI inside an Aion project, or as a standalone Node package installed from npm. ## Composer Controls | Control | Behavior | | ------------- | ------------------------------------------------ | | `Enter` | Send the message or select the active menu item. | | `Shift+Enter` | Insert a newline. | | `@` | Open the agent picker. | | `/` | Open the slash command picker. | | `Esc` | Dismiss the active menu or clear the draft. | | `Ctrl+C` | Clear the draft or exit when the draft is empty. | ## Python SDK CLI Use the Python SDK CLI when you are working inside a Python SDK project or want to launch chat from `poetry run aion chat`. Install it from the [Python SDK installation guide](/sdk/python/installation). Start interactive chat: ```bash theme={null} poetry run aion chat ``` By default, `aion chat` connects to `http://localhost:8000`. That default is useful when you are running a local Aion proxy or agent server with the Python SDK. ### Options | Option | Default | Description | | ----------------------- | ----------------------- | ------------------------------------------------------------------- | | `--url`, `--host`, `-u` | `http://localhost:8000` | Agent or proxy URL to connect to. | | `--agent-id` | none | Agent identifier for proxy-aware routing with an explicit endpoint. | | `--token` | none | Bearer token for authenticated explicit endpoints. | | `--header` | none | Custom HTTP header in `key=value` format. Repeatable. | | `--push-notifications` | disabled | Include local push notification configuration. | | `--push-receiver` | `http://localhost:5000` | Push notification receiver URL. | | `--help` | n/a | Show help text. | `aion chat run` also supports headless one-shot messages: | Option | Default | Description | | -------------------- | -------------- | --------------------------------------------------------------- | | `--agent` | none | Discovered agent handle, display id, identity id, or agent key. | | `--request-mode` | `send-message` | `send-message` or `streaming-message`. | | `--response-mode` | `message` | `message` for rendered output or `a2a` for raw protocol JSON. | | positional `message` | none | Message text. Use `-` to read stdin. | ### Examples Connect to the default local endpoint: ```bash theme={null} poetry run aion chat ``` Target a different endpoint: ```bash theme={null} poetry run aion chat --url http://localhost:8000 ``` Target a specific agent through a proxy: ```bash theme={null} poetry run aion chat --url http://localhost:8000 --agent-id support ``` Send a headless message to a discovered agent: ```bash theme={null} poetry run aion chat run --agent @david "Hello" ``` Write raw A2A protocol output for a headless message: ```bash theme={null} poetry run aion chat run --agent @david --response-mode a2a "Hello" ``` See [Command Line Interface](/sdk/python/running/cli) for the broader Python SDK CLI. ## Standalone npm Package Use the standalone npm package when you want the terminal chat client without installing the Python SDK. Install the standalone chat tool globally: ```bash theme={null} npm install -g @terminal-research/aion ``` Run the primary executable: ```bash theme={null} aio ``` By default, `aio` connects to `http://localhost:8000`. The package also installs an `aion-chat` alias: ```bash theme={null} aion-chat ``` ### Options | Option | Default | Description | | ----------------------- | ----------------------- | ------------------------------------------------------------------- | | `--url`, `--host`, `-u` | `http://localhost:8000` | Agent or proxy URL to connect to. | | `--agent-id` | none | Agent identifier for proxy-aware routing with an explicit endpoint. | | `--token` | none | Bearer token for authenticated explicit endpoints. | | `--header` | none | Custom HTTP header in `key=value` format. Repeatable. | | `--push-notifications` | disabled | Include local push notification configuration. | | `--push-receiver` | `http://localhost:5000` | Push notification receiver URL. | | `--help` | n/a | Show help text. | | `--version` | n/a | Print the package version. | `aio run` and `aion-chat run` also support headless one-shot messages: | Option | Default | Description | | -------------------- | -------------- | --------------------------------------------------------------- | | `--agent` | none | Discovered agent handle, display id, identity id, or agent key. | | `--request-mode` | `send-message` | `send-message` or `streaming-message`. | | `--response-mode` | `message` | `message` for rendered output or `a2a` for raw protocol JSON. | | positional `message` | none | Message text. Use `-` to read stdin. | ### Examples Connect to the default local endpoint: ```bash theme={null} aio ``` Connect through a local proxy and select an agent: ```bash theme={null} aio --url http://localhost:8000 --agent-id onboarding ``` Send an authenticated request: ```bash theme={null} aio --url https://agent.example.com --token "$AION_TOKEN" ``` Add custom headers: ```bash theme={null} aio --header "X-Client=aion-docs" --header "X-Trace=demo" ``` Enable push notifications: ```bash theme={null} aio --push-notifications --push-receiver http://localhost:5000 ``` Send a headless message to a discovered agent: ```bash theme={null} aio run --agent @david "Hello" ``` Write raw A2A protocol output for a headless message: ```bash theme={null} aio run --agent @david --response-mode a2a "Hello" ```