What a connector actually does for you
You could do all of this with secrets — generate an API key, paste it in, ask the agent to write the calls. A connector removes the fiddly parts:- You don’t handle a token. Sign in through the service and Paige stores what it needs.
- Your bot gets a ready-made helper, so the agent writes
postMessage(...)rather than hand-rolling HTTP calls against an API it’s half-remembering. - The agent knows the service. It looks up the real reference before writing anything, including your actual Airtable tables and fields.
What’s available
Airtable
Read and write records in an Airtable base — look up an order, log a booking, update a status.
Slack
Post messages to a Slack channel — alert your team when someone needs a human.
Conversations
Conversations is also listed here, but it’s a different sort of thing. There’s no account to sign in to and no token — clicking Connect simply asks the Code agent to add a small piece of code to your bot, and that’s what lets you reply to customers from the Conversations tab. It only appears on Paige-hosted projects. Once it’s in, the card reads Active.Connecting
Each card offers up to two routes:- Connect with service — the one-click route. A popup opens, you sign in and approve, and it closes itself.
- Or enter an access token manually — paste a token you’ve made yourself. Useful when your organisation restricts app installs, or you want to control exactly which permissions the token carries.
Connectors and deploy
Your connection is shared between your preview bot and your live bot, and doesn’t need a deploy — connect once and both can use it.One exception: Airtable’s Refresh schema writes an updated list of your tables and fields into your bot code, which does need a deploy to reach your live bot. See the Airtable page.
