The API for asking humans

Give your AI agent a voice.

Your agent can talk to every system you own — but the humans who hold the answers it needs are usually outside its chat window. SurVoyce is the human-I/O layer: your agent creates a voice form, hands a person a link or a phone number, and reads back clean, structured answers. No human operator required.

MCP nativeA2A readySelf-serve keysHumans answer by voice

# watch an agent onboard itself — every line below is real

$ curl -X POST https://api.survoyce.com/public/agents/register \

-H 'content-type: application/json' -d '{"name": "acme-research-agent", "organization": "Acme Research"}'

{"workspaceId": "…", "apiKey": "sv_…", "trialCredits": 3, "verified": false}

$ # connect over MCP and build a form

$ mcp → https://api.survoyce.com/api/mcp/survoyce/mcp

tools: create_form · add_question · publish_form · list_submissions · get_answers …

$ # a human answers by voice, then:

$ get_answers submission_id=…

{"nps_score": 9, "churn_reason": null, "quote": "honestly, keep the phone option"}

# time-to-first-key: one curl▌

Jump to the quickstart

How it works for agents

1. Register

One unauthenticated POST returns an API key and trial credits. No OAuth dance, no human sign-up flow.

2. Build & publish

Create forms, add typed questions with validation, and publish — over MCP tools or plain REST.

3. Hand a human the mic

Every published form gets a web link and a phone form code. The person answers in a natural voice conversation — with a clear AI disclosure up front, always.

4. Read the answers

get_answers returns a keyed JSON object — {"email_address": …, "overall_rating": …} — plus transcripts where retained.

Quickstart

1. Get a key

curl -X POST https://api.survoyce.com/public/agents/register \
  -H 'content-type: application/json' \
  -d '{"name": "acme-research-agent", "organization": "Acme Research"}'

The apiKey is shown once — store it. Rotate any time via POST /v1/keys/rotate.

2. Connect over MCP

{
  "mcpServers": {
    "survoyce": {
      "url": "https://api.survoyce.com/api/mcp/survoyce/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Streamable HTTP at /api/mcp/survoyce/mcp, SSE at /api/mcp/survoyce/sse. Prefer A2A? The intake agent's card is at /api/.well-known/intake/agent-card.json, execution at /api/a2a/intake.

What you can do — and the limits

Create, edit, and publish voice forms; list submissions; read keyed JSON answers; buy credits programmatically (Stripe).
New agent workspaces start unverified: trial credits and reduced daily caps apply until verification.
Outbound calling is unavailable to unverified workspaces. Respondents come to you via your form's link or phone code.
Every session opens with a mandatory AI disclosure to the human. It cannot be disabled — by you or by us.

Are you an agent? Skip the prose:

register:   POST https://api.survoyce.com/public/agents/register
auth:       Authorization: Bearer <apiKey>
mcp:        https://api.survoyce.com/api/mcp/survoyce/mcp  (SSE: /api/mcp/survoyce/sse)
a2a card:   https://api.survoyce.com/api/.well-known/intake/agent-card.json
a2a exec:   POST https://api.survoyce.com/api/a2a/intake
key rotate: POST https://api.survoyce.com/v1/keys/rotate
openapi:    https://api.survoyce.com/openapi.json
docs:       https://survoyce.com/docs/agents-api.md  (index: https://survoyce.com/llms.txt)
humans:     answer by voice; AI disclosure is mandatory; abandoned sessions cost nothing