Skip to main content
Connecting a webhook source is a short sequence of API calls: create a source, mint an ingest token, point the provider’s webhook at Splendor, then confirm events arrive. An AI coding agent (Claude Code, Codex, and similar) can run that sequence for you — you tell it which provider you want and approve the steps.

The prompt

Paste the prompt below into your coding agent and tell it which provider to connect (for example, “GitHub”, “Stripe”, “Linear”, or “Sentry”). Fill in the three values first.

What the agent does

The prompt drives the same flow as Connect a source:
1

Creates a source

Calls POST /v1/admin/hosted-sources for a dataset you name, and saves the source_id and source_key.
2

Mints an ingest token

Calls POST /v1/admin/sources/{source_id}/ingest-token and keeps the plaintext token (returned once) to use as the webhook’s bearer credential.
3

Wires up the webhook

Gives you a small relay that wraps each provider event as a one-element batch and forwards it to /v1/ingest/sources/{source_key}/logs with the ingest token, then has you point the provider’s webhook at the relay. (If the provider can send a JSON array with a custom Authorization header, it skips the relay and posts straight to the logs endpoint.)
4

Verifies and searches

After a test event, checks GET /v1/ingest/runs for the dataset and runs a search to confirm the events are queryable.
The prompt instructs the agent to show you each request before it sends it and to wait for your go-ahead on any write.

Secrets stay out of chat

The ingest token is shown once when minted — treat it like a password. Have the agent store it where the provider (or your relay) reads it, not in the chat transcript. The agent never needs your provider credentials: you configure the webhook in the provider’s own dashboard.
Prefer to run the steps yourself? Connect a source lists the same calls as copy-paste curl commands.