Edit branding
Set the agent's caller-facing branding on `agents.settings.branding`: `displayName` (the business name shown on the request_details_form detail-capture page the caller fills in, as the website chat widget's header name, in the header of the agent's public share page, and on the staff transcript link of a website chat handoff) and `formMessage` (a short line under it on that form, used nowhere else). In the app both live under Business details on the agent's Channels tab. Pass only the field you want to change; an empty string or null clears it (the display name then falls back to the agent's name). Set this to the customer's trading name, a leftover onboarding placeholder is caller-visible whenever the agent texts a details form. The logo is uploaded separately and is not settable here.
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_branding" \ -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
}Update behaviour POST
Update conversation + turn-detection + context-memory + transfer-detection knobs, maxResponseTokens (50-2000 spoken-reply brevity in TOKENS, spoken as roughly tokens × 0.75 words, cut at a sentence boundary once a turn runs well past that, NOT a token cap), llmTemperature (0-1), llmThinking (the model reasons before each reply; only on the models that offer thinking), llmEffort (the thinking level while llmThinking is true: low, medium or high), backgroundNoise (ambience bed under the call), backgroundNoiseVolume (0-1 loudness of that bed), backgroundNoiseProfile (which bed plays; default office), typingSpinner (a sound during a tool lookup), typingSpinnerSound (which clip plays; default keyboard), pauses (the named pauses the prompt writes as [pause:NAME]; replaces the whole list), shortTurnTimeout (0.5-5 seconds), longTurnTimeout (15-30 seconds; floored at 15), eotThreshold (0-1 end-of-turn confidence), eagerEotThreshold (0-0.9 quick-response eagerness; higher = replies sooner), autoSummarization, maxContextTokens (1000-16000), maxUnsummarizedMessages (5-100), summaryTargetTokens (500-8000), amdEnabled, amdTier2Enabled. shortTurnTimeout and longTurnTimeout are in SECONDS; eotThreshold/eagerEotThreshold are 0-1 confidences; AMD stands for answering-machine detection. Send only what changes. Use this for 'shorter answers' (maxResponseTokens down), 'less likely to interrupt' (eotThreshold up), 'more creative' (llmTemperature up), 'think before answering' (llmThinking true), 'think harder' (llmThinking true with llmEffort high), 'answer faster' on a model that offers thinking (llmThinking false), 'wait longer before responding' (longTurnTimeout up), 'add background noise' (backgroundNoise true), 'quieter background' (backgroundNoiseVolume down), 'make it sound like a cafe' (backgroundNoiseProfile cafe), 'play a typing sound during lookups' (typingSpinner true), 'use the typewriter sound' (typingSpinnerSound typewriter), 'add a 1.5 second typing pause called lookup' (pauses with { name: lookup, seconds: 1.5, sound: keyboard } added to the existing list), 'summarise long calls' (autoSummarization true), 'detect voicemail before transferring' (amdEnabled). Pass null for a field to unset it, back to the default.
Save brief POST
Store the agent's brief, the two interview answers that can be checked rather than just written down. `callReasons` is what people ring about. `nevers` is what the agent must never do, and each never must name the tool it forbids, because storing it also COMPILES it: the platform creates a scenario where a caller pushes three times for the forbidden thing, asserts `tool_not_called` against that tool, and links the scenario to the rule. A never that names no tool, or names a tool this version does not have, is REFUSED and nothing is stored, so ask the operator which action they mean instead of guessing, and read the tool list off get_agent rather than from memory. Removing a never from the list deletes its scenario too. Call this during the interview, before writing config: the answers persist on the draft version, so they survive into the next build and a later reader can see what the operator actually asked for. This records intent, it does not change how the agent behaves. Wire the behaviour itself with update_system_prompt and the matching set_* ops.