Skip to main content
Describe what you want in plain words, and Reagent reads your REAPER project, runs the tools it needs and tells you what it did. To read or change your project, it needs the Bridge, a small script that runs inside REAPER (see Quick Start).

​
Sending Messages

Type your request and press Enter to send, or Shift+Enter for a new line. The chat box starts in the middle of a new chat and moves to the bottom once the chat has messages. You can be conversational or direct. Both of these work:
You can also ask about your project, or about files in folders Reagent can reach, such as your home folder and external drives (see Where Reagent Can Reach):
Specific requests work better than vague ones. For tips and more examples, see Best Practices. In the chat box you can also:
  • Run a saved prompt. Type / to pick a saved command. See Slash Commands.
  • Paste long text. Pastes over 10 lines collapse into a “Script pasted · N lines” block that you can expand or remove before you send. A paste too large to send is turned away with a message, not cut short.
Reagent can make mistakes. Back up your REAPER projects before you ask for big changes.

​
Reference Selections

Every time you send a message, Reagent reads your selected items and tracks, edit cursor position and time selection in REAPER, plus the Wwise selection if Wwise is connected. Just refer to your selection. Reagent can also look up other parts of the project, such as regions, when you mention them.

​
Following Reagent’s Work

Each run of tool steps folds into a summary line. Short notes Reagent writes between runs appear in lighter text, and the next run gets its own line.
  • While it works, the line names the step in progress (“+N more” when several run at once) and shows a timer, such as “Running for 12s”.
  • When it’s done, the line sums up the steps, such as “Read the tracks” or “Ran 6 tools”, and the reply ends with “Worked for 1m 35s”. Time spent waiting for your answer isn’t counted.
  • Click the summary line to see each step, with its details and results.
Reagent starts with a core set of tools and loads more, or a skill, when a task needs them. These Load tools and Load skill steps don’t change anything in REAPER, Wwise or your files. Before a step that changes your project or files, Reagent may ask first on an approval card. The permission selector next to + sets how often it asks: Default permissions, Auto-review or Full access, and you can change it while Reagent works. See Permissions for what each mode asks about. If a step fails, the summary line says so, for example “Ran 6 tools (1 failed)”. Open it to see:
  • A one-line description of the problem, in red or grey depending on how serious it is.
  • Sometimes, a suggested next step marked with →.
  • Show details for the technical error.
A failed step doesn’t end the reply. Reagent sees the error and can try a different approach. If it keeps failing, stop the reply and describe the task another way.
When a whole reply fails, you see one of these messages. For more fixes, see Chat Problems.

​
Answering Reagent’s Questions

When Reagent needs a decision, a question card replaces the chat box. If the Reagent window isn’t in front, you get a desktop notification: “Reagent needs your input”. With several questions, the card shows “Question 1 of 3” and steps through them.

​
Cancelling an Agent Question

  • Click the X at the top of the card.
  • On single-choice questions, click Cancel next to Something else.
  • Press Esc. This cancels the question and stops the reply. While you’re typing in Something else, Esc clears your text instead.
Reagent stops instead of guessing. The chat shows “You cancelled — type what you’d like to do”, and the cursor moves to the chat box so you can redirect it.

​
Stopping a Reply

While Reagent works, the send button becomes a square stop button (Cancel request). Click it or press Esc. You can’t type in the chat box while Reagent works, so stop the reply to change course.
  • The reply is labelled “Cancelled”, and its summary line adds “(cancelled)”.
  • Stopping doesn’t undo changes Reagent already made. Check your project before you continue, and see Undoing Changes to reverse them.
  • Stopping doesn’t give credits back, including for the step that was running. See Plans and Credits.

​
Attaching Files

Attach files to give Reagent something to work from, such as a screenshot of a plugin, a CSV of new file names or a PDF brief. You can:
  • Click + in the chat box and choose Attach file, Attach image or Attach folder.
  • Drag files or folders onto the chat box.
  • Paste an image, such as a screenshot, from the clipboard.
  • Paste a file’s full path on its own, such as /Users/you/Desktop/brief.pdf or D:\Projects\brief.pdf.
Attachments appear as tiles above your text. Click a tile’s X to remove it. You can send attachments without any text. An attached folder appears as a chip. Reagent gets its path, not its contents, and can use its file tools on the files inside if the folder is within reach. See Where Reagent Can Reach. You can attach up to 10 files per message, and up to 8 MB per image, text file or PDF. Files turned away, for example because of their type or size, are reported in a pop-up and aren’t added.
If the file exists and is supported, Reagent attaches it. GIF, HEIC and HEIF images are turned away, and any other path stays as text.Use the full path, without quotation marks (on Windows, remove the ones Copy as path adds). Shortcuts such as ~/ aren’t recognized.
Attach the script as a PDF, TXT, CSV, TSV, Word or Excel file and mention it in your request. See Dialogue.
Your attachments plus the conversation so far must fit within the size limit of each request to Reagent’s service. Reagent warns you when you attach a file that would go over. If earlier attachments have made the conversation too large, it warns you before sending and lets you remove one.If a file fails to upload to Reagent’s servers, the chat box shows “name could not be attached. Remove it to send this message.” The message won’t send until you remove the file.

​
Where Attachments Go

  • Individual plans: attachments go to the AI model through Reagent’s service. PDF, TXT, CSV, TSV, Word and Excel files are also uploaded to Reagent’s servers when you attach them, so the dialogue tools can use them as a reference script. See Documents You Attach.
  • Enterprise accounts: images, PDFs and text files go to your organization’s model. Reagent doesn’t upload reference files to its servers for Enterprise accounts, so Word and Excel files are turned away. To work from a spreadsheet, save it as CSV or TSV.

​
Managing Chats

Reagent opens on a new chat. Earlier chats are in the sidebar, grouped into Starred, Today, Yesterday, Previous 7 days and Older.
  • New chat: click New chat at the top of the sidebar, or press Cmd+N (macOS) or Ctrl+N (Windows). Start one for each unrelated task, so Reagent stays focused and sends less to the model with each message. A new chat has no memory of earlier chats or the results gathered in them, such as analysis, so finish a multi-step job before you switch.
  • Open: click an earlier chat.
  • Search: type in Search chats to filter by title.
  • Star: hover a chat and click the star icon, or use its ⋯ menu. Starred chats stay at the top.
  • Rename or edit: Reagent names each chat automatically. You can’t rename a chat or edit a sent message. Send a new message instead.
  • Delete: in the ⋯ menu, choose Delete (one chat) or Clear all (every chat), then confirm. You can’t restore a deleted chat in the app.
  • Hide the sidebar: press Cmd+B (macOS) or Ctrl+B (Windows).
Where chats are saved. Reagent saves your chats automatically on this computer. They don’t sync to your account or to other computers, and signing out doesn’t delete them. Anyone who opens Reagent on this computer with the same computer login (your macOS or Windows user account) can see them, even after signing in with a different Reagent account. On a shared computer, use your own computer login for Reagent. See Privacy & Security.
Run one chat at a time. Reagent lets you send in a second chat while the first is still working, but both would change the same REAPER project at once. Wait for the reply to finish, or stop it, before you work in another chat.

​
Long Chats and the Context Ring

Once Reagent has replied, a small ring next to the send button shows how much of the model’s context window (how much of the conversation it can take in at once) is used. Hover over it to see “N% context used”. It turns amber at 80% and red at 95%. When a chat gets long, Reagent automatically summarizes older turns for the model and adds a divider: “Earlier context summarized … Everything above is still here.” You still see the whole chat, but Reagent may no longer recall fine details from far back. If that matters, start a new chat. On Enterprise accounts, both follow the Context window set for your organization’s endpoint.

​
Copying a Chat

Click Copy chat in the title bar to copy the whole conversation as Markdown, for example to share a workflow or report a problem. It appears once the chat has messages. To copy one of your messages, hover over it and click Copy message.

​
Sending Feedback

To report a bug or suggest a feature, use the option for your plan:
  • Individual plans: click the Send Feedback icon in the title bar, or ask Reagent in chat to pass a note on to the team. For the steps and what’s sent, see Send Feedback From the App.
  • Enterprise accounts: use the contact page.

Permissions

Choose when Reagent asks before it changes your project, and undo changes.

REAPER Tools

Everything Reagent can do in your REAPER project.

Best Practices

Write clear requests and keep your projects safe.

Keyboard Shortcuts

Every shortcut in Reagent.