Limits and what is sent
Messages answered by your own model use no Captain credits, but a few limits still apply. This page lists them and says what Captain sends to your provider.
Daily message limit
Section titled “Daily message limit”There is a limit on how many messages your own model can answer per day.
- The day is counted in UTC, so it resets at midnight UTC, not at your local midnight.
- The limit is for your account: chats about all your profiles count together.
When you reach it, Captain shows a message that names the limit: “You’ve used today’s … messages with your own model. Switch to Captain credits, or try again tomorrow (UTC).” Set the switch in the message box to Credits to carry on that day.
Other limits with your own model
Section titled “Other limits with your own model”- No credit budget. A message on your own model has no credit cap, because it uses no credits. The limit on steps per message still applies: “Captain reached its step limit for one message. Ask it to continue.”
- Model checks. Check and save and Check again together are limited to 10 in 10 minutes.
- Test call time. The check waits up to 45 seconds for your endpoint to answer.
- Reply size. A single reply from your endpoint can be at most 8 MB.
- No redirects. Captain does not follow a redirect from your endpoint. Enter the final address.
- Images. Image attachments are offered only when Reads images is on for your model.
The general limits apply as well: one message at a time, changes in batches, and the five-minute wait on confirmations. See Limits.
What is sent to your endpoint
Section titled “What is sent to your endpoint”When your model answers, Captain sends the following to the endpoint you entered, using your API key:
- Your message.
- The content of the profile the chat is about: an outline of its sections and entries, and the sections Captain reads while it works.
- The recent messages and replies of the conversation, and a short summary of older ones.
- The text read from documents you attached, in this message or earlier in the conversation.
- Images you attach, when Reads images is on.
- Captain’s own instructions and the list of actions it can take.
Your provider handles that content under its own terms, not under the Captain terms. The Model page asks you to confirm the address before it saves.
Nothing from your other profiles is sent.
What Captain keeps
Section titled “What Captain keeps”- Your API key, encrypted. It is used only for your Captain messages and is never shown again. Remove my model deletes it.
- The conversation, as for any Captain chat. See What is sent to the model.
- A record of each message in your history, marked Your own model, with no cost.
When your provider has a problem
Section titled “When your provider has a problem”Errors from your provider are shown as yours to fix, for example:
- “Your model provider rejected your API key. Check it in Model settings.”
- “Your model provider says the model you chose doesn’t exist.”
- “Your model provider is rate limiting requests. Try again later.”
- “Captain could not reach your model endpoint.”
The latest error also appears on the Model page under Your model had a problem. Fix the cause at your provider, then select Check again. See Set it up.
If you see “Your model isn’t available”, your model is not connected or your plan no longer includes the feature. Switch to Credits or connect the model again.