Set tool outputs
Keep fields of an in-call HTTP tool's response as variables for after the call. `outputs` maps a snake_case variable name to a dot path in the JSON response, the same grammar as a workflow function's outputs: {"booking_id": "booking.id", "first_slot": "slots[0].start"}. After each successful response the runtime reads each path and keeps the value for the rest of the call; a repeat call keeps the latest value, and a missing path or a non-JSON body keeps nothing and never changes what the agent hears. Hang-up and post-call functions then read the value as {{booking_id}}. A name must not be a system or post-call variable, a caller detail, a post-call field or a pre-call function output. Replaces the whole map; pass {} to keep none. Nothing is kept when the agent stores basic details only.
Authorization
bearerAuth A personal access token from Settings > API tokens.
In: header
Header Parameters
A key you choose, such as a UUID, that makes this POST safe to retry. Within 24 hours the same key and request answers with the first successful answer and runs nothing, the same key with a different request answers 422 idempotency_key_reused, and a key whose request is still running answers 409 idempotency_key_in_use. A failed request keeps nothing, so the same key runs it again.
1 <= length <= 255Request Body
application/json
TypeScript Definitions
Use the request body type in TypeScript.
Response Body
application/json
application/problem+json
application/problem+json
application/problem+json
application/problem+json
application/problem+json
application/problem+json
application/problem+json
curl -X POST "https://example.com/operations/set_tool_outputs" \ -H "Content-Type: application/json" \ -d '{ "agent": "string" }'{}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}{
"type": "http://example.com",
"title": "string",
"status": 0,
"detail": "string",
"code": "string",
"errors": [
{
"path": "string",
"message": "string"
}
],
"estimate": {},
"retryAfterSeconds": 0
}Set tool mock POST
Configure mock mode on an HTTP tool. Identify the tool by toolName (its snake_case name) or toolId. Mock mode lets the tool return a canned response, `mockResponse`, after `mockDelayMs`, instead of calling its real URL, so an agent can be tested before its API exists and without firing real side effects. Mocks are scoped to test surfaces: on a phone call carrying a real caller the runtime refuses to serve the canned body and returns an error envelope instead, so never leave a mock on as a substitute for a working endpoint. Pass only the fields you want to change; omit the rest. Builtin tools can't be mocked. To 'edit the mock response', send `mockResponse` with the new JSON string. A mock body may echo the call's own arguments with `{{args.field}}`, write `{"category": "{{args.category}}"}` rather than a fixed value for anything the real API would reflect back, or the mock will contradict what the agent sent and a judge will fail the run for it. TIME THIS EDIT: a mock lives on the agent's DRAFT, so changing one starts a new version and every test green on the old one goes stale (get_test_results marks them). Do it at the start of a draft cycle, not while re-greening a suite you are about to publish. To make a mocked verify_* tool reject a wrong identifier, send `mockMatch`, a JSON string of {arg: expected-value} pairs the call must match (a value can nest an object or array for a tool whose argument is itself structured). To mock a tool that returns DIFFERENT data for different inputs (order A → record A, order B → record B, else → not found), send `mockMocks`, a JSON string of [{match, response, delayMs?}, …]; it supersedes the single mock when set.
Set tool params POST
Replace an HTTP tool's input parameter schema. Identify the tool by toolName (its snake_case name) or toolId. `properties` is a map of name → `{ type, description }`. `required` is the list of names the LLM MUST provide. Names may use only letters, digits, `_`, `.` and `-`, up to 64 characters, or the write is refused. The runtime injects these into the request body (POST) or query string (GET). Optionally pass `input_examples`, worked example calls the runtime forwards to the model as its `input_examples` field, for any tool whose arguments are nested, optional, or format-sensitive. This REPLACES the whole schema, so re-send `input_examples` when editing `properties` or they are dropped.