Menu

Board AI chats and credits | Rodeo Docs

A chat on a board is an AI assistant that reads only the cards you wire into it, so you can ask "give me ten hooks from these interviews" and get an answer grounded in your own material, with the cost shown before you send.

The "Campaign angles" chat open in focus mode in a side panel: the four wired cards shown as chips at the top, the prompt chips below, and the message box with "≈ N credits" underneath
The "Campaign angles" chat open in focus mode in a side panel: the four wired cards shown as chips at the top, the prompt chips below, and the message box with "≈ N credits" underneath

Do it

Start a chat

Press C, or click New chat in the dock. A chat card appears on the board.

Wire cards into it: drag from the handle on the right of each card to the chat. See Nodes and wires.

The empty chat tells you how many cards it can read, for example "Ask about the 4 connected cards.", and offers Saved prompts as chips.

Type your question in the box, which says "Ask anything, or type / for prompts", and press Enter. Shift+Enter adds a new line.

The reply streams in live, for you and anyone else looking at the board. The wires into the chat light up while it writes. Click Stop (the square) to stop a reply early.

Give it room

Click Open in side panel (the arrows at the top right of the chat) for a full-height panel; × puts it back. Or select a chat and press F for focus mode, which also lists the wired cards as chips at the top.

Choose a model and thinking

In the chat's title row:

Model: Sonnet (the default), Haiku (fast) or Opus.

Thinking (the brain icon): turns on extended thinking, for questions that need more reasoning.

The small ring shows how full the chat's context is: everything wired in, plus the conversation so far. Hover it for the exact number of tokens.

You can also click the microphone in the message box to Dictate a question instead of typing it.

Use the reply

Under each reply:

A chat in the side panel with a question and its reply, and the row of actions under the reply highlighted: Send to board, Send selection to board, Copy, Retry, Save as page, Save as script, Add selection to script and Save as new version
A chat in the side panel with a question and its reply, and the row of actions under the reply highlighted: Send to board, Send selection to board, Copy, Retry, Save as page, Save as script, Add selection to script and Save as new version

Action

What it does

Send to board

Puts the reply on the board as a note beside the chat.

Send selection to board

Select part of the reply first; just that part becomes a note.

Copy

Copies the reply.

Retry

Asks again for a new answer (newest reply only).

Save as page

Saves the reply as a page under the board. See Turn a board into a page.

Save as script, Add selection to script, Save as new version

Write into a script. See Scripts and editor notes.

Credits

Every AI reply on a board uses AI credits from your workspace's monthly allowance. The bigger the context, the longer the reply and the larger the model, the more it uses.

Under the message box, ≈ N credits is the most the next message can cost. It's an upper bound, so you usually pay less.

After a reply, Last reply N credits shows what it actually cost.

When the workspace has none left, chats show You're out of AI credits this month. Owners and admins see Upgrade; everyone else sees Ask an admin. Click Try again after credits are added.

A reply that has started always finishes, even if it takes the workspace past its allowance. The next message is then refused until credits reset.

Plan

AI credits per month

Opus

Free

150 for the workspace

No

Starter

100 per member

No

Team

200 per member

Yes

Business

350 per member

Yes

Founders' Deal

200 per member

Yes

Credits reset at the start of each calendar month (UTC). Research chats, Write up as page and editor notes use credits too. Dictation uses source minutes instead: see Sources. Check your workspace's balance in Usage.

The copilot

The Copilot button at the top right of every board opens an assistant that changes the board for you. Ask it to "group the YouTube sources and wire them into a new chat" and it adds, moves, groups and wires cards. Anything it wants to remove waits for you: click Allow or Deny on each change, or Allow all for this run. Revert run undoes what one request did. Each request counts as an agent run and uses AI credits.

Tips and limits

A chat reads exactly its wires, nothing cut short. If the context ring fills up, sending is blocked with Disconnect or shorten sources: remove a wire or two. Sources must show a ✓ to be read.

"Sources changed since this reply" appears under the last reply when its wired cards or wires changed. Ask again for a fresh answer.

One reply at a time per chat. A message can be up to 50,000 characters.

Private material stays private. A reply built from a page you can't open shows "This reply used pages you can't open".

View-only? You can read every chat but not send. The box says "View only".

Ideas

Hooks in one minute. Wire your customer interviews and a competitor scan into a chat, pick 10 hook angles, and send the best three to the board as notes.

A cheap first pass, a careful second. Draft with Haiku (fast), then switch to Sonnet with Thinking on for the final outline.

Two chats, one board. Wire the same sources into an "Instagram" chat and a "Newsletter" chat, and get both versions side by side.

Troubleshooting

"Opus is available on the Team and Business plans". Your workspace is on Free or Starter. Pick Sonnet or Haiku (fast), or upgrade. See Plans and limits.

The reply ignores a card. Check the card has a wire into this chat and a ✓. Press F to see the wired cards listed.

Related