> ## Documentation Index > Fetch the complete documentation index at: https://app.miyon.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Create a voice agent > Create a voice agent with its prompt. Optionally set prompt, voice and knowledge-base values in the same request; they are validated exactly as in the settings endpoints and the whole request fails if any value is rejected. Add tools afterwards with `POST /api/v2/tools` using the returned `id` as `agent_id`. ## OpenAPI ````yaml /api-reference/openapi.yaml post /api/v1/voice_agents openapi: 3.1.0 info: title: Customer API version: 1.0.0-beta description: > **Beta.** This API is in beta. Contracts can still change while we complete it, including breaking changes; we announce changes in the release notes and keep them as rare as we can. Pin your integration to the field names and status codes documented here and treat unknown fields as optional. REST API to create and configure voice agents, add tools, test them with evals, read conversations, and place AI outbound calls. **Authentication.** Voice agents, settings, tools, evals and conversations use an API token created on the **Keys** page of your dashboard. Each operation lists the scope it needs under `x-required-scopes`; `:*` grants every action of a resource. Outbound calls use a separate outbound API token from the **Outbound Calls** page. Send either token as `Authorization: Bearer `. **Responses.** A single resource is returned as a flat JSON object; a collection is wrapped in a plural key (`voice_agents`, `tools`, `evals`, `conversations`) next to its paging fields. Deleting returns `204 No Content`. Everything is addressed by the voice agent's `id`; prompts and numbers are internal. **Errors.** Authentication failures return `{ "error", "message" }`. Request quotas return `429` with a `Retry-After` header and `{ "error", "message", "details" }`, where `error` is a stable machine-readable code. Every API call must finish within 30 seconds; longer requests return `503` with `mf-core:api-request-timeout`. Validation errors list the problems per field; the exact shape is shown on each operation. servers: - url: https://app.miyon.ai security: [] tags: - name: Voice Agents description: >- Create and manage voice agents. Configure them through the voice, prompt and knowledge-base settings endpoints. - name: Voice Agent Settings description: >- Read and change a voice agent's voice, prompt and knowledge-base settings. A voice agent has one prompt (its instructions, greeting and conversation settings). Read the schema of a settings group, then update one `key`/`value` pair per request. - name: Tools description: >- Create, configure, enable and execute agent tools. Tools are enabled per voice agent with `agent_id`. These endpoints accept scoped API tokens only; every error uses the `{ "error", "message", "details" }` envelope with a stable `mf-core:tool-*` code. - name: Outbound Calls description: >- Place AI calls and review their outcome. Authenticate with the outbound API token from the **Outbound Calls** page. - name: Conversations description: >- List and inspect the conversations of your account, including transcripts when your retention settings allow. - name: Evals description: >- Evals are automated test conversations that check how a voice agent responds. A scenario contains cases; each case contains the caller's turns and what the agent must say or do. Run a scenario to get pass/fail results. paths: /api/v1/voice_agents: post: tags: - Voice Agents summary: Create a voice agent description: > Create a voice agent with its prompt. Optionally set prompt, voice and knowledge-base values in the same request; they are validated exactly as in the settings endpoints and the whole request fails if any value is rejected. Add tools afterwards with `POST /api/v2/tools` using the returned `id` as `agent_id`. operationId: createVoiceAgent requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/VoiceAgentCreateRequest' example: name: Booking agent description: Takes appointment calls prompt: user_prompt: You help callers book appointments. initial_response: Hello, how can I help you? voice_config: voice_config.response_time_ms: 1200 responses: '201': description: The created voice agent. content: application/json: schema: $ref: '#/components/schemas/VoiceAgent' '401': $ref: '#/components/responses/Unauthorized' '403': $ref: '#/components/responses/Forbidden' '422': description: > Nothing was created. `{ errors: { name: ["is required"] } }`, `{ errors: { tools: [...] } }`, `{ errors: { prompt | voice_config | knowledge_base: ["unsupported keys: ..."] } }`, `{ errors: { base: [...] } }` from agent creation, or the concern envelope with `errors` when a nested value was rejected. content: application/json: schema: oneOf: - $ref: '#/components/schemas/FieldErrors' title: Field errors - $ref: '#/components/schemas/ConcernResponse' title: Settings response with errors example: errors: name: - is required '429': $ref: '#/components/responses/RateLimited' '503': $ref: '#/components/responses/ServiceUnavailable' security: - scopedJwt: [] components: schemas: VoiceAgentCreateRequest: type: object description: > Flat JSON object. Unknown top-level keys are ignored. The nested sections are validated exactly as in the settings endpoints; if any value is rejected, nothing is created (HTTP 422). A `tools` key is rejected: create and enable tools with the Tools endpoints after the agent exists. properties: name: type: string minLength: 1 description: Display name of the agent. description: type: string description: Free-text description shown in the dashboard. prompt: type: object description: >- Prompt settings. Keys may also be given in dotted form (`prompts.user_prompt`). Unknown keys return 422. properties: user_prompt: type: string description: The agent's instructions. initial_response: type: string maxLength: 1000 description: The greeting. mute_user_while_assistant_speaks: type: boolean transcription_locale: type: string description: >- A supported locale code from the prompt schema, or `multi` for automatic language detection. voice_config: type: object description: >- Dotted voice settings keys as listed by `GET /api/v1/voice_agents/{id}/concerns/voice_config/schema`, for example `voice_config.response_time_ms`. Unknown keys return 422. propertyNames: pattern: ^voice_config\. additionalProperties: true knowledge_base: type: object properties: selected_datasource_ids: type: array items: type: integer description: >- IDs of the knowledge sources the agent may use. Replaces the current selection. description: Knowledge sources to attach to the agent's prompt. required: - name VoiceAgent: type: object properties: id: type: integer description: ID of the voice agent. name: type: - string - 'null' description: Display name of the agent. description: type: string description: Empty string when no description is set. created_at: type: string format: date-time description: When the agent was created. updated_at: type: string format: date-time description: When the agent or its prompt was last changed. concerns: type: object description: URLs of this agent's settings endpoints. properties: voice_config: $ref: '#/components/schemas/ConcernLinks' prompts: $ref: '#/components/schemas/ConcernLinks' knowledge_base: $ref: '#/components/schemas/ConcernLinks' required: - voice_config - prompts - knowledge_base required: - id - name - description - created_at - updated_at - concerns FieldErrors: type: object description: Field-keyed error envelope. `base` holds errors not tied to one field. properties: errors: type: object additionalProperties: type: array items: type: string description: Problems per field; `base` for problems not tied to one field. required: - errors additionalProperties: false ConcernResponse: type: object description: >- Response format for reading and updating a settings group. On 409 and 422 the same format is returned with `errors` filled and `values`/`revision` reflecting the unchanged current state. properties: concern: type: string enum: - voice_config - prompts - knowledge_base description: The settings group this response belongs to. shape: type: string const: blocks description: Layout hint for clients that render the settings; always `blocks`. schema_version: type: string examples: - voice-config.v1 - prompts.v4 - knowledge_base.v2 description: Version of the settings schema, for example `voice-config.v1`. save_mode: type: 'null' description: Reserved; always null. revision: type: string description: >- Current revision. Send it as `If-Match` on voice settings writes to avoid overwriting concurrent changes. values: type: object additionalProperties: true description: Dotted field key to current value. schema: type: 'null' description: Reserved; always null. Read the schema from the schema endpoint. schema_revision: type: 'null' description: Reserved; always null. effects: type: array items: $ref: '#/components/schemas/ConcernEffect' description: Related settings that changed as a consequence of this update. errors: type: object additionalProperties: type: array items: type: string description: Problems per field key. Empty on success. draft: type: 'null' description: Reserved; always null. affected_concerns: type: array items: type: object properties: concern: type: string revision: type: string required: - concern - revision description: Settings groups whose revision changed. system_prompt: type: string description: >- Voice settings updates only. Present when the update regenerated the agent's instructions. required: - concern - shape - schema_version - revision - values - effects - errors - affected_concerns ConcernLinks: type: object properties: read_url: type: string description: GET the current values. schema_url: type: string description: GET the fields and accepted values. change_url: type: string description: PATCH one `key`/`value` pair. required: - read_url - schema_url - change_url description: Read, schema and update URLs of one settings group. AuthError: type: object description: >- Returned when the API token is missing, invalid, expired or lacks the required scope. properties: error: type: string description: Short error category. examples: - Authentication failed - Forbidden message: type: string description: Human-readable explanation. Absent on some 403 responses. examples: - 'Missing required scope: voice_agents:write' required: - error ConcernEffect: type: object description: A dependent value that changed as a side effect of the write. properties: key: type: string description: Dotted key of the setting that changed. old_value: description: Value before the update. value: description: Value after the update. reason: type: - string - 'null' description: Why the setting changed, when the server can say. required: - key ApiError: type: object description: >- Error format with a stable machine-readable `error` code, a human-readable `message` and request-specific `details`. properties: error: type: string description: >- Stable machine-readable error code, for example `mf-core:rate-limit-exceeded`. message: type: string description: Human-readable explanation. details: type: object additionalProperties: true description: >- Request-specific data, for example the offending parameter or the quota policy. required: - error - message - details responses: Unauthorized: description: Missing, expired or invalid bearer token. content: application/json: schema: $ref: '#/components/schemas/AuthError' example: error: Authentication failed message: Missing bearer token Forbidden: description: >- The token does not carry the scope this operation requires (see `x-required-scopes`). content: application/json: schema: $ref: '#/components/schemas/AuthError' example: error: Forbidden message: 'Missing required scope: voice_agents:write' RateLimited: description: >- Too many requests. Wait the number of seconds given in `Retry-After` before retrying. headers: Retry-After: description: Seconds to wait before retrying. schema: type: integer minimum: 1 content: application/json: schema: $ref: '#/components/schemas/ApiError' example: error: mf-core:rate-limit-exceeded message: Rate limit exceeded details: policy: api_write retry_after: 3 ServiceUnavailable: description: >- The request was not completed. Either it exceeded the 30-second request time limit that applies to every API call (narrow the request and retry), or rate limiting is temporarily unavailable. Retry shortly. content: application/json: schema: $ref: '#/components/schemas/ApiError' examples: timeout: value: error: mf-core:api-request-timeout message: The API request exceeded the 30-second time limit details: {} limiter: value: error: mf-core:rate-limiter-unavailable message: API rate limiting is temporarily unavailable details: policy: api_write securitySchemes: scopedJwt: type: http scheme: bearer bearerFormat: JWT description: >- Scoped API token. Required scopes are listed on each operation in x-required-scopes. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.