Quickstart

This page walks you through the golden path — every step takes ≤ 60 seconds. All Agntix APIs are served from a single base URL behind the API gateway:

https://api.agntix.ai

The gateway routes traffic to the appropriate upstream service based on the path prefix. As a customer you only ever talk to api.agntix.ai — internal service URLs are not exposed.

Create an account

Get your API key

Open Settings → API Keys, click Create key, pick Read & write, and copy the value. It looks like:

pk_live_xxxxxxxxxxxxxxxxxxxxxxxx

Set it in your shell:

export AGNTIX_API_KEY=pk_live_xxxxxxxxxxxxxxxxxxxxxxxx

Keys grant full access to your organization. Store them in a secret manager — never commit one to git.

Make your first authenticated call

GET /v1/chat/models lists every model your org can use. It needs auth but no body — perfect for a smoke test.

curl https://api.agntix.ai/v1/chat/models \
  -H "x-api-key: $AGNTIX_API_KEY"

const res = await fetch("https://api.agntix.ai/v1/chat/models", {
  headers: { "x-api-key": process.env.AGNTIX_API_KEY },
});
const { data } = await res.json();
console.log(data.length, "models available");

import os, httpx
res = httpx.get(
  "https://api.agntix.ai/v1/chat/models",
  headers={"x-api-key": os.environ["AGNTIX_API_KEY"]},
)
print(len(res.json()["data"]), "models available")

A 200 OK here means your key works.

Create an agent

The minimum agent needs a name, a model, and a system prompt.

curl https://api.agntix.ai/v1/chat/agents \
  -H "x-api-key: $AGNTIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Quickstart bot",
    "modelId": "gpt-4o-mini",
    "systemPrompt": "You are a friendly assistant. Be concise."
  }'

const res = await fetch("https://api.agntix.ai/v1/chat/agents", {
  method: "POST",
  headers: {
    "x-api-key": process.env.AGNTIX_API_KEY,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    name: "Quickstart bot",
    modelId: "gpt-4o-mini",
    systemPrompt: "You are a friendly assistant. Be concise.",
  }),
});
const agent = await res.json();
console.log(agent.id);

res = httpx.post(
  "https://api.agntix.ai/v1/chat/agents",
  headers={"x-api-key": os.environ["AGNTIX_API_KEY"]},
  json={
    "name": "Quickstart bot",
    "modelId": "gpt-4o-mini",
    "systemPrompt": "You are a friendly assistant. Be concise.",
  },
)
print(res.json()["id"])

Save the id from the response — you’ll need it next.

Send a message

Open a session against the agent, then post a user message. The response includes the assistant’s reply.

SESSION=$(curl -s https://api.agntix.ai/v1/chat/chat/sessions \
  -H "x-api-key: $AGNTIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"agentId\": \"$AGENT_ID\"}" | jq -r .id)

curl https://api.agntix.ai/v1/chat/chat/sessions/$SESSION/messages \
  -H "x-api-key: $AGNTIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"content": "Say hello in three words."}'

const session = await (await fetch("https://api.agntix.ai/v1/chat/chat/sessions", {
  method: "POST",
  headers: { "x-api-key": process.env.AGNTIX_API_KEY, "Content-Type": "application/json" },
  body: JSON.stringify({ agentId }),
})).json();

const reply = await (await fetch(
  `https://api.agntix.ai/v1/chat/chat/sessions/${session.id}/messages`,
  {
    method: "POST",
    headers: { "x-api-key": process.env.AGNTIX_API_KEY, "Content-Type": "application/json" },
    body: JSON.stringify({ content: "Say hello in three words." }),
  },
)).json();

console.log(reply);

session = httpx.post(
  "https://api.agntix.ai/v1/chat/chat/sessions",
  headers={"x-api-key": os.environ["AGNTIX_API_KEY"]},
  json={"agentId": agent_id},
).json()

reply = httpx.post(
  f"https://api.agntix.ai/v1/chat/chat/sessions/{session['id']}/messages",
  headers={"x-api-key": os.environ["AGNTIX_API_KEY"]},
  json={"content": "Say hello in three words."},
).json()
print(reply)

You’re done

You now have:

An agent your team can iterate on from the dashboard or the API.
A session with a real LLM-generated reply.
A working API key suitable for your local dev environment.

What’s next

Add a knowledge store (RAG)

Make the agent answer from your own documents instead of generic web knowledge.

Wire up voice

Same agent, now speaking — over the browser or PSTN phone calls.

Subscribe to webhooks

Push session lifecycle events to your back-office.

Production checklist

Rate limits, retries, error handling, and SLOs.

Getting Started

Build

Real-time

SDKs & Examples

Operate

You’re done

What’s next

Add a knowledge store (RAG)

Wire up voice

Subscribe to webhooks

Production checklist

​You’re done

​What’s next

Add a knowledge store (RAG)

Wire up voice

Subscribe to webhooks

Production checklist

You’re done

What’s next