> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.jambonz.org/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.jambonz.org/_mcp/server. # Result webhook Every call's result appears in [Calls](/agent-studio/calls). If your campaign system needs it too, Studio can post it to a URL of yours: who answered, whether the goal was reached, how a transfer went, the call's length and its ids. ## Turn it on The result webhook is set on an outbound agent, in its **Placing calls** section. At the bottom of the **Placing outbound calls** card, open the fold that reads **Results shown in Calls here**, then: 1. Turn on **Also send each call's result to your own webhook**. 2. Enter the **Result webhook URL**. It must be a public `https` address, without a user name or password in it. The URL belongs to the agent, not to a version: it takes effect for the next call that ends, without publishing. Turning the switch off stops deliveries. ## When it fires Studio posts once per call, shortly after the call ends and Studio has its complete record. It fires for every call the agent handles while the URL is set, including test calls; `is_test` tells them apart. ## The request ```http POST /your/path HTTP/1.1 content-type: application/json user-agent: studio-webhook/1 x-studio-event: call.completed x-studio-delivery: x-studio-signature: t=1761742800,v1=5f0c… ``` `x-studio-delivery` is the call id. It is the same on every retry of a delivery, so you can use it to ignore duplicates. ## The payload ```json { "event": "call.completed", "call_id": "b4d1c3e2-…", "agent": "0193f5a2-…", "version": 4, "direction": "outbound", "answered_by": "person", "amd_ms": 1850, "goal_reached": true, "goal_reason": "tool:book_appointment", "transfer": { "result": "none", "mode": "", "ring_ms": null }, "tools": [{ "name": "book_appointment", "ok": true, "ms": 412 }], "interruptions": 1, "turns": 7, "latency_ms": { "p50": 980, "max": 2140 }, "completion_reason": "normal", "call_termination_by": "caller", "duration_s": 94, "call_status": "completed", "started_at": "2026-10-29T14:02:11.204Z", "ended_at": "2026-10-29T14:03:47.950Z", "is_test": false, "sip_callid": "8a7e…@10.0.1.5", "experiment": null, "transcript": [ { "role": "assistant", "text": "Hi, this is Sam from Acme Dental." }, { "role": "user", "text": "Hi Sam." } ] } ``` | Field | What it is | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------ | | `event` | always `call.completed` | | `call_id` | the jambonz call id | | `agent` | the agent's id | | `version` | the version number that handled the call | | `direction` | `inbound` or `outbound` | | `answered_by` | `person`, `machine`, `no_answer`, `undetermined` or `unknown` | | `amd_ms` | how long answering machine detection took, or `null` | | `goal_reached` | `true` or `false`, or `null` when the agent has no goal | | `goal_reason` | what reached the goal: `transfer_bridged`, `tool:` and the tool's name, or `model_marked`; empty when not reached | | `transfer` | `result` is `none`, `bridged`, `returned` or `failed`; `mode` is how the transfer was made; `ring_ms` is how long it rang | | `tools` | each tool call: its name, whether it succeeded, and its round-trip time in milliseconds | | `interruptions` | how many times the caller cut the agent off | | `turns` | the number of turns | | `latency_ms` | the caller's wait per turn: the median (`p50`) and the slowest (`max`) | | `completion_reason` | why the agent's session ended, as jambonz reports it | | `call_termination_by` | who ended the call, as jambonz reports it | | `duration_s` | seconds from answer to end; `0` when the call was not answered | | `call_status` | the call's final status from jambonz, such as `completed`, `no-answer`, `busy` or `failed` | | `started_at`, `ended_at` | ISO 8601 times; `ended_at` can be `null` | | `is_test` | `true` for a test call | | `sip_callid` | the SIP Call-ID | | `experiment` | `{ "id", "group" }` when the call was in an [experiment](/agent-studio/experiments), with `group` `A` or `B`; otherwise `null` | | `transcript` | the conversation, as `role` (`user` or `assistant`) and `text`; present only when the transcript was kept | A payload larger than 256 KB is not sent; the delivery history shows it as failed with `too_large`. ## Verify the signature Each delivery is signed with the agent's **signing secret**. The `x-studio-signature` header has two parts: * `t`: the time of signing, in Unix seconds * `v1`: the HMAC-SHA256, in hex, of the string `.`, keyed with the signing secret To verify a delivery, compute the same HMAC over the raw request body exactly as received, before parsing it, and compare. Rejecting deliveries whose `t` is too old protects you from replays. ```js import { createHmac, timingSafeEqual } from 'node:crypto'; // rawBody: the request body as a string, exactly as received export function verifyStudioSignature(rawBody, header, secret, toleranceS = 300) { const parts = Object.fromEntries( header.split(',').map((p) => { const i = p.indexOf('='); return [p.slice(0, i), p.slice(i + 1)]; }), ); const t = Number(parts.t); if (!Number.isFinite(t) || !parts.v1) return false; if (Math.abs(Date.now() / 1000 - t) > toleranceS) return false; const expected = createHmac('sha256', secret).update(`${parts.t}.${rawBody}`).digest('hex'); const a = Buffer.from(expected, 'hex'); const b = Buffer.from(parts.v1, 'hex'); return a.length === b.length && timingSafeEqual(a, b); } ``` ### The signing secret Under **Signing secret**, below the URL: * **Reveal** shows the secret, for one minute. It starts with `whsec_`. Use the whole string as the HMAC key. * **Rotate** replaces it. Deliveries from then on are signed with the new secret and the old one stops working at once, so update your receiver first. Revealing and rotating the secret are limited to Owners and Admins, and both are recorded in the account's audit log. ## Delivery and retries A delivery succeeds when your endpoint answers with a `2xx` status within 10 seconds. | Your endpoint | What Studio does | | ------------------------------------------------------------- | -------------------------------------------------------------------------- | | answers `2xx` | delivered | | answers `5xx`, `408` or `429`, times out, or can't be reached | tries again, up to 3 more times: after 5 seconds, 30 seconds and 5 minutes | | answers another `4xx` | gives up; no retry | | answers with a redirect | gives up; redirects are not followed | | resolves to a private address | gives up; the delivery is blocked | If 25 calls in a row fail to deliver, Studio pauses the agent's webhook for an hour. Calls that end during the pause are not delivered; each one is recorded in the delivery history with the error `paused`. Deliveries resume by themselves after the hour. Studio sends at most 3 deliveries at a time for each account; others wait their turn. ## Delivery history **Recent deliveries**, below the signing secret, lists the latest 20 delivery attempts for the agent, newest first: | Column | What it shows | | ----------- | ------------------------------------------------------------------------------------------------------------------------------ | | **When** | when the attempt was made | | **Call** | the start of the call id | | **Attempt** | 1 for the first try, then each retry; 0 for a call that was not sent (paused, or too large) | | **Status** | **Delivered**, the HTTP status your endpoint returned, or **Failed** | | **Error** | why it failed, for example `http 500`, `timeout`, `redirect not followed`, `blocked: private address`, `paused` or `too_large` | While deliveries are paused, the list says so at the top. Recent deliveries is shown to Owners and Admins. ## Next * [Publishing](/agent-studio/publishing): place calls from an outbound agent * [Calls](/agent-studio/calls): every call's result, in Studio