Who answers my visitors when codus is closed?
When your Mac is closed or codus is not running, codus can keep answering visitors from the cloud. It replies from a short "support brief" of your product that you generate from the bound project. It answers only from that brief and never has your code, so it stays safe. If a message looks like a bug, it raises an issue and still gives the visitor a helpful reply. If it cannot answer safely, it parks the conversation and tells the visitor the team will get back to them. When you reopen codus, it quietly takes those conversations back.
Before you start
- The instance must be bound to a project. See How do I connect a support chat to a project?.
- After-hours answering needs a support brief, which you generate from the instance card. No brief means no after-hours answers.
Generate a support brief
The brief is a short, plain-language description of your product, distilled from the bound project's code. It is what the cloud answers from when it cannot read the code directly.
- Open the chats tab and click ▶ instances.
- Expand the instance and find the support-brief line.
- Click generate brief (or regenerate brief if one already exists). It reads "distilling…" while it works, which takes a few minutes.
The instance card shows the brief's age, for example "support brief: fresh today" or "support brief: 5d old". If there is none, it reads "no support brief yet — the cloud responder needs one to answer while you are away". Keep the brief reasonably current: an old brief is treated as untrustworthy and the cloud will hold back rather than risk answering from stale information.
What after-hours answering will and will not do
- It answers only from the brief and the conversation itself. It never has access to your code or systems when it runs in the cloud.
- It will not invent product behaviour. If the brief does not cover something, it says the team will confirm.
- It does not make promises about pricing, refunds, or anything legal.
- If a message is bug-shaped (something is broken, an error, lost data), it raises an issue for your team and still replies to the visitor to say it has been logged. That issue is waiting for you in the issues tab when you return.
A reply that came from the cloud shows in your chats tab with an "agent · cloud" chip, so you can tell at a glance which answers went out while you were away.
When it parks instead of answering
Sometimes the safe thing is to not answer. In two cases the cloud parks the conversation rather than reply:
- No fresh brief. Without an up-to-date brief it has nothing reliable to answer from.
- The daily limit is reached. Each instance has a small daily cap so after-hours answering can never run up an unbounded bill.
When it parks, the visitor sees a message like "Thanks for the message — the team has been notified and will reply here as soon as possible", and the conversation shows as waiting for you. Nothing is lost; it is held for a person.
Coming back online
When you reopen codus and sign in, your desktop takes over coverage again. Conversations the cloud was holding come back to your chats tab and floor, and from then on they are answered with full access to the project's code once more. You never have to hand anything back manually.
How do I know it worked?
Generate a brief, then close codus and send a test message through the widget on your site. Within a minute or two you should get a reply in the widget, and when you reopen codus that conversation is in your chats tab with an "agent · cloud" chip on the reply.
Common problems
Nothing answers after hours. The most common cause is a missing or stale support brief. Open the instance and click generate brief (or regenerate brief). See the brief section above.
Visitors get "the team has been notified" and no real answer overnight. The conversation was parked, either because the brief is not fresh or the daily limit was reached. Regenerate the brief; the parked conversations are waiting as waiting in your chats tab.
The answers are thinner after hours than when I am online. That is expected. Online, the agent can read your code directly; after hours it works from the brief only. A fuller, fresher brief gives better after-hours answers.