REST endpoints

Send a message

Send a customer message to an agent and get its answer — the same answer the chat bubble would give. Optionally streamed.

POST/api/v1/chat

Send your API key in the Authorization: Bearer af_live_… header. See Authentication.

Body

agent_idstringrequired

The agent's ID (where to find it).

messagestringrequired

The customer's message, 1–4,000 characters.

user_idstring

Your own ID for this person (max 200 characters). Messages with the same user_id continue the same conversation.

conversation_idstring

Continue a specific conversation returned earlier. It must belong to this agent and have been created through the API.

namestring

The customer's name, saved on a new conversation.

emailstring

The customer's email, saved on a new conversation.

verified_emailstring

An email address your server vouches for. The conversation counts as a verified customer, so actions limited to verified customers can run for this email.

streamboolean

true to receive the answer as Server-Sent Events. Default false.

Without conversation_id or user_id, every request starts a new conversation.

Request

curl https://anytimereply.com/api/v1/chat \
  -H "Authorization: Bearer af_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id": "AGENT_ID",
    "message": "Do you deliver to Sylhet?",
    "user_id": "customer-123",
    "name": "Rafi",
    "email": "[email protected]"
  }'

Response

200 OK

{
  "conversation_id": "8b1d6c2e-…",
  "message_id": "f04a…",
  "text": "Yes! We deliver across Bangladesh, including Sylhet, in 2–4 days. Delivery is ৳120.",
  "sources": [{ "title": "Shipping & delivery", "url": "https://example.com/shipping" }],
  "confidence": 0.82,
  "handoff": false,
  "waiting_for_human": false,
  "ui": null
}

Response fields

conversation_idstring

Keep it (or keep using your user_id) to continue the conversation.

message_idstring | null

The stored reply. null when no reply was written (see below).

textstring

The agent's answer. Plain text with light Markdown (links, bold, lists).

sourcesarray

Up to 3 pages the answer came from (title, url). Empty when Show sources is off.

confidencenumber | null

How well your knowledge matched the question (0–1). null for Q&A, flows and small talk.

handoffboolean

true when this message handed the chat to your team.

waiting_for_humanboolean

true when your team has taken over this conversation. The AI doesn't answer; text is empty.

uiobject | null

Buttons, forms, tap-to-reply choices and widgets to show under the answer: buttons (id, type, label, url), form, replies (strings) and widgets.

When your team takes over

After a handoff the conversation appears in your Helpdesk inbox. Team replies are stored in the conversation, but the API can't push them to you — read them with GET /api/v1/conversations/{id} (messages with role: "human") or listen for the handoff.requested webhook.

If the chat was paused by Guardrails (spam or a blocklist), text is empty and message_id is null.

Streaming

With "stream": true the response is text/event-stream. Each event is one data: line with JSON:

Text

data: {"type":"meta","conversation_id":"8b1d6c2e-…"}

data: {"type":"token","t":"Yes! We "}

data: {"type":"token","t":"deliver across Bangladesh…"}

data: {"type":"done","message_id":"f04a…","text":"Yes! We deliver across Bangladesh…","sources":[…],"handoff":false,"waiting_for_human":false,"ui":null}

Event	Fields
`meta`	`conversation_id` — sent first
`token`	`t` — the next piece of text
`done`	The full answer: `message_id`, `text`, `sources`, `handoff`, `waiting_for_human`, `ui`
`error`	`error` — a readable message; the stream then ends

Read the stream (Node.js 18+)

const res = await fetch("https://anytimereply.com/api/v1/chat", {
  method: "POST",
  headers: { Authorization: "Bearer " + process.env.ANYTIMEREPLY_API_KEY, "Content-Type": "application/json" },
  body: JSON.stringify({ agent_id: "AGENT_ID", message: "Hi!", user_id: "customer-123", stream: true }),
});
const reader = res.body.getReader();
const decoder = new TextDecoder();
let buffer = "";
for (;;) {
  const { done, value } = await reader.read();
  if (done) break;
  buffer += decoder.decode(value, { stream: true });
  const events = buffer.split("\n\n");
  buffer = events.pop();
  for (const e of events) {
    if (!e.startsWith("data: ")) continue;
    const msg = JSON.parse(e.slice(6));
    if (msg.type === "token") process.stdout.write(msg.t);
    if (msg.type === "done") console.log("\n", msg.sources);
  }
}

The final text in done can contain more than the streamed tokens (for example links added after an action) — use it as the source of truth.

Notes

Conversations from the API appear in Chat logs with the channel API and fire the conversation.created webhook.
Each answer uses message credits like the chat bubble.
Answers can take a few seconds (longer when actions run); allow up to 120 seconds before timing out.

Errors

Status	Error
`400`	`conversation_id not found for this agent` · validation errors such as `message: Required`
`401`	Missing or invalid API key
`404`	`Agent not found`
`429`	Rate limit (120 requests/minute per key)

Was this page helpful?

Still stuck? Contact us