> ## Documentation Index
> Fetch the complete documentation index at: https://docs.alpha.new/llms.txt
> Use this file to discover all available pages before exploring further.
# The composer
> Type, attach, dictate and mention: everything you can put in a message to the agent.
The composer is the box at the bottom of a session where you write to the agent: text, files, dictation, mentions and context picked from the workspace. Sending while the agent is still working has its own page, [Queue, steer or stop](https://docs.alpha.new/chat/queue-and-steer.md). For the composer before a session exists, see [The home screen](https://docs.alpha.new/sessions/home-screen.md).
Settings sit on the left of the composer. Voice and send sit on the right.
| Control | What it does |
| - | - |
| **+** (**Add Attachment**) | Uploads files into the session's machine |
| Model chip | Shows the model, plus the reasoning effort once you've picked one. Click it, or type `/model`, to [change them](https://docs.alpha.new/chat/agents-and-models.md). |
| **plan** chip | Opens a picker that turns [plan mode](https://docs.alpha.new/chat/plan-mode.md) **On** or **Off**. The chip reads **plan · on** while it's on. ⇧Tab switches it from the keyboard. |
| Microphone | Dictates into the composer |
| **↑** (**Send message**) | Sends the message |
| Square stop button (**Stop generating**) | Replaces send while the agent works and the composer is empty. It stops the turn, as Esc does. |
While the agent works and you've typed something, the send button queues or steers instead. Its tooltip names your default, such as "Queue · ⌘⏎ to steer".
## Send a message
Press Enter to send and ⇧Enter for a new line. A session still called "New session" takes its name from your message; `/rename` changes it later.
The placeholder tells you what the next message will do:
| Placeholder | When you see it |
| - | - |
| "Describe what to build — / for commands, @ for files…" | The agent is idle, plan mode is off and the session isn't a routine |
| "Describe what to plan — the agent explores and asks, then proposes a plan…" | Plan mode is on |
| "What should run, and when? The agent builds, tests and schedules it…" | The session is a [routine](https://docs.alpha.new/routines.md) |
| "Add a follow-up" | The agent is working, so your message queues or steers |
In a narrow composer the longer placeholders are cut to their first half, such as "Describe what to build…".
One message can be up to [100,000 characters](https://docs.alpha.new/reference/limits.md). A longer one is refused with "prompt too large", so attach long text as a file. If a send fails, the banner says why and your text goes back into the composer.
## Attach files
Click **+** or drag files onto the composer. Each file uploads as soon as you attach it, not when you send, into `/workspace/uploads/` on the session's machine, where the agent opens it with its ordinary tools. Attaching wakes a sleeping machine.
Your message gains one line per file, such as `[Attached file: uploads/harbor-classes.csv (12.4 KB)]`. Images up to [15 MB](https://docs.alpha.new/reference/limits.md) also reach the model as pictures, so a screenshot of a bug or a design works. A larger image arrives only as a file.
| You do | What happens |
| - | - |
| Attach a file whose name is taken | Alpha renames it, so `report.md` becomes `report-1.md` |
| Send before an upload finishes | The message shows at once with "Sending once the upload finishes…" and goes when the file lands |
| Click an image tile | Opens the image larger |
| Click a tile's X (**Remove file**) | Removes the tile and deletes the file from the machine |
Each file can be up to [256 MB per file (default)](https://docs.alpha.new/reference/limits.md). If an upload goes wrong, you see one of these. [Troubleshooting](https://docs.alpha.new/reference/troubleshooting.md) covers the rest.
| Message | Meaning |
| - | - |
| A red alert icon on the tile | The upload failed. Hover the tile for the reason ("Upload failed" when there is none), then remove it and attach the file again. |
| "An attachment failed to upload — the message wasn't sent." | A file that was still uploading when you sent failed. Your text is back in the composer. |
| "The session was switched before the upload finished — the message wasn't sent." | You sent, then opened another session before a file finished uploading. The message goes back into its own session's composer, so go back and send it again. |
Sent files show as tiles above your message in the transcript. [Files and uploads](https://docs.alpha.new/workspace/files.md) covers the `uploads` folder.
## Mention files, sessions and skills
Three characters open a list above the composer. Use ↑ ↓ to move, Enter to pick, and Esc to close.
| Type | Lists | Notes |
| - | - | - |
| `/` at the start of the message | Commands | Opens only when the message is the command. Codex commands Alpha doesn't run yet show **not in Alpha yet**. See [Slash commands](https://docs.alpha.new/reference/slash-commands.md). |
| `@` | Files and folders in the session's workspace | Up to 50 matches, with the matching letters highlighted. Picking one inserts `@path`, drawn as a badge. |
| `@session:` and one word | Your other sessions whose messages or title contain that word | Up to 15 sessions. Picking one inserts a link the agent can open and read. |
| `$` | The skills the agent can use | Codex sessions. Picking one inserts `$skill-name`. |
| `/` after other text | The skills the agent can use | Claude Code sessions, where a skill runs as `/name` |
You can also drag a file or folder from the file tree onto the composer to mention it. Backspace at the end of a mention removes all of it. [Skills](https://docs.alpha.new/skills/overview.md) explains what a skill adds.
### Point the agent at another session
Type `@session:` and a word from the session you mean, such as `@session:booking`. Each hit shows the session's title, the date it was last active and a line from where the word matched. The search takes one word: a space after it closes the list. The session you're in is never listed.
## Dictate a message
Click the microphone, or tap ⌘., to start dictating. Tap again to stop. To talk only while you hold the keys, hold ⌘. and let go when you're done. Use Ctrl instead of ⌘ on Windows and Linux.
Speak any language; Alpha writes it in English. Text you type while dictating is never overwritten. After you stop, the button spins and its tooltip reads "Finishing…" while the last words settle. A dictation runs for up to [2 minutes](https://docs.alpha.new/reference/limits.md), then stops by itself.
| Message | Fix |
| - | - |
| "Microphone access was blocked — allow it in the browser to dictate." | Allow the microphone for Alpha in your browser's site settings |
| "Voice input disconnected — press the mic to try again." | Click the microphone again |
| "Voice input can't use this microphone's sample rate (…)" | Pick another microphone in your system or browser settings |
> [!NOTE]
>
> If you don't see the microphone, voice input isn't turned on for your Alpha, or your browser doesn't support it.
## What happens to my unsent text?
Alpha keeps your unsent text in this browser, one draft per session. Sending clears it.
| Event | Your draft |
| - | - |
| You reload, close the tab or switch sessions | The text is kept. Attachments are not. |
| The same session is open in two tabs | Both share one draft, and the last tab you typed in wins |
| The text looks like a key | It isn't saved to the browser at all |
| You sign out of Alpha | Every draft in this browser is cleared |
When the composer holds something shaped like a known key, a warning appears above your text: "That looks like an OpenAI API key. Keys typed here land in the transcript — add it under Secrets instead, and the agent gets it as an environment variable." A GitHub token points you to the **GitHub** tab instead. The warning never blocks sending.
> [!WARNING]
>
> A key you send stays in the session's transcript, readable by anyone you [share the session](https://docs.alpha.new/sharing.md) with. Alpha hides saved secrets' values in the transcript, but a key typed into a message isn't a saved secret. Add keys under [Secrets](https://docs.alpha.new/secrets.md) instead.
## Pick up a suggested next prompt
Claude Code only
After a turn, a Claude Code session may show a chip above the composer holding the prompt Claude Code predicts you'll want next. Hover it to read the whole prompt. A click replaces anything in the composer and sends the suggestion at once, so read it first. Codex sessions don't show these chips.
## Add context from the workspace
Selections and picks from the workspace wait above your text as chips: a selection in the code editor or a viewer (**Add to prompt**), an element picked in the Preview with **Select**, or the Preview's logs (**Attach the last lines to your next message**). Each chip has an X to remove it. On send, each chip's context goes ahead of your text, so the agent knows exactly which lines or element you mean. Marked-up pictures from **Add to chat** arrive as image attachments. [Point at and draw on your app](https://docs.alpha.new/workspace/point-and-draw.md) shows every source.
## Next steps
> **[Queue, steer or stop](https://docs.alpha.new/chat/queue-and-steer.md)**
>
> Send while the agent works: wait your turn, join the running turn, or stop it.
> **[Point at and draw on your app](https://docs.alpha.new/workspace/point-and-draw.md)**
>
> Pick an element in the Preview or mark up a screenshot, and it lands in the composer.
> **[Slash commands](https://docs.alpha.new/reference/slash-commands.md)**
>
> Every command in the `/` list, for Codex and Claude Code sessions.
> **[Secrets](https://docs.alpha.new/secrets.md)**
>
> Where API keys belong, so they never land in the transcript.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.