Configuration
- URL – the endpoint to call. Supports environment variable substitution with
${VAR_NAME}syntax, and prompt inputs can fill in its path or query with{{placeholders}}, e.g./users/{{user_id}}. A password or a credential query parameter –api_key,tokenand the like – takes its secret from a placeholder, e.g.?api_key=${API_KEY} - Method – GET, POST, PUT, PATCH, DELETE
- Headers – custom headers, also supporting
${VAR_NAME}substitution. Credential headers –Authorization,Cookie, and any header named like a key, token or secret – take their secret from a placeholder, e.g.Authorization: Bearer ${API_TOKEN} - Parameters – defined as a JSON Schema. For GET requests, parameters are serialised into the query string. For other methods, they travel as a JSON request body
Environment variable substitution
Use${VAR_NAME} in URLs and headers to inject values from your workspace’s environment variables at runtime. This keeps secrets out of your workflow definitions – store an API key as an environment variable, reference it in the header, and the actual value gets injected when the tool runs.
Correlation headers
Every request the tool makes carries two headers identifying the run it came from, so your endpoint can correlate calls back to a PromptJuggler run:PromptJuggler-RunId– the root run ID. It is shared across every prompt and workflow in the same run tree, so nested tool calls all report the same value.PromptJuggler-Thread– the conversation thread ID.
Limits
- Responses up to 256 KiB – a longer response reaches the model cut at 256 KiB, with a note saying so.
- Redirects – followed up to 10 times. Your custom headers go only to the host you configured and its subdomains, so a redirect elsewhere, such as to a CDN or a presigned storage URL, never sees your keys.