Use Kimi K3 Open API from your server

Quick start

The API key is shown once after creation. Copy it immediately, store it in server-side configuration, and never place it in client-side JavaScript, mobile app bundles or public repositories.

export K3NOVA_API_KEY="<k3nova-server-api-key>"

Connect K3Nova to Buzz

Buzz users can run K3Nova as a channel agent without sharing their Buzz private key with k3nova.com. The customer creates or chooses a Buzz agent identity inside their own Buzz community, adds that agent to the target channel, creates a K3Nova server API key, then runs the bridge from a private machine or server.

export BUZZ_RELAY_URL="https://your-buzz-relay.example"
export BUZZ_PRIVATE_KEY="<customer-buzz-agent-private-key>"
export BUZZ_CHANNEL_ID="your-channel-id"
export K3NOVA_API_KEY="$K3NOVA_API_KEY"

npm run buzz:bridge

The bridge listens for @K3Nova in the selected channel, sends the thread context to /api/v1/chat/completions, and posts the answer back to the same Buzz thread. The machine-readable setup manifest is available at /api/integrations/buzz/manifest.

List available models

Use the models endpoint to confirm that your key works before sending a chat request.

curl https://k3nova.com/api/v1/models \
  -H "Authorization: Bearer $K3NOVA_API_KEY"

Create a chat completion

For a simple JSON response, set stream to false. Streaming responses are also supported for server clients that can read event streams.

curl https://k3nova.com/api/v1/chat/completions \
  -H "Authorization: Bearer $K3NOVA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "stream": false,
    "temperature": 1,
    "top_p": 1,
    "max_completion_tokens": 512,
    "presence_penalty": 0,
    "frequency_penalty": 0,
    "stop": ["END"],
    "n": 1,
    "user": "server-user-123",
    "messages": [
      {
        "role": "user",
        "content": "Write a concise launch checklist for a new API integration."
      }
    ]
  }'

JavaScript fetch example

This example is for Node.js or another private server runtime. Keep K3NOVA_API_KEY outside browser code.

const response = await fetch("https://k3nova.com/api/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.K3NOVA_API_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "kimi-k3",
    stream: false,
    temperature: 1,
    top_p: 1,
    max_completion_tokens: 512,
    messages: [
      { role: "user", content: "Summarize this support ticket in three bullets." }
    ]
  })
});

if (!response.ok) {
  throw new Error(await response.text());
}

const data = await response.json();
console.log(data.choices?.[0]?.message?.content || data);

Chat parameters

The chat endpoint accepts the common OpenAI-style controls below. Unknown fields are ignored by the gateway.

Errors and limits

Each key has its own daily Agent credits cap. API calls also use the account plan allowance, so a request can stop when the key, plan, rate gate, provider configuration or remaining allowance is not ready.