Skip to content
Kindo Kindo Docs

Responses API quickstart

POST https://api.kindo.ai/v1/responses speaks the OpenAI Responses API. Stock OpenAI clients work unmodified. Pick the section below that matches the client you already use.

Prerequisites

codex

codex is the OpenAI CLI built on the Responses API. Point it at Kindo with two environment variables:

Terminal window
export OPENAI_BASE_URL=https://api.kindo.ai/v1
export OPENAI_API_KEY=$KINDO_API_KEY


codex

That’s the entire integration. codex builds standard Responses requests and Kindo handles them with stock OpenAI semantics.

openai-python

import os
from openai import OpenAI


client = OpenAI(
    base_url="https://api.kindo.ai/v1",
    api_key=os.environ["KINDO_API_KEY"],
)


response = client.responses.create(
    model="claude-sonnet-4-5-20250929",
    input="Explain Kubernetes pod security policies in plain English.",
)


print(response.output_text)

openai-node

import OpenAI from 'openai';


const client = new OpenAI({
  baseURL: 'https://api.kindo.ai/v1',
  apiKey: process.env.KINDO_API_KEY
});


const response = await client.responses.create({
  model: 'claude-sonnet-4-5-20250929',
  input: 'Explain Kubernetes pod security policies in plain English.'
});


console.log(response.output_text);

curl

Terminal window
curl -X POST https://api.kindo.ai/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $KINDO_API_KEY" \
  -d '{
    "model": "claude-sonnet-4-5-20250929",
    "input": "Explain Kubernetes pod security policies in plain English."
  }'

Response shape

The response uses the standard OpenAI Responses output:

{
  "id": "resp_abc123",
  "object": "response",
  "status": "completed",
  "model": "claude-sonnet-4-5-20250929",
  "output": [
    {
      "type": "message",
      "role": "assistant",
      "content": [
        {
          "type": "output_text",
          "text": "Pod Security Policies were Kubernetes rules that controlled what a pod was allowed to do..."
        }
      ],
      "status": "completed"
    }
  ],
  "output_text": "Pod Security Policies were Kubernetes rules that controlled what a pod was allowed to do..."
}

Raw HTTP responses also include output_text at the top level — the SDK exposes it via a convenience field of the same name.

What you get by default

  • A single-shot, stateless completion. Nothing is persisted on Kindo’s side, and previous_response_id / conversation carry no state across calls.
  • Your instructions (or input-array developer / system messages) flow through verbatim. Kindo does not prepend its curated system prompt.
  • Tools are only what you list in tools. Kindo does not silently inject kindo_shell, kindo_web_search, or anything else.

If you want any of those — stateful conversations, a curated Kindo system prompt, or the Kindo-hosted tool catalog — opt in via the Chat Actions extensions.

Next steps

  • Request shape — every standard Responses field Kindo honors and how it’s forwarded.
  • Streaming — set stream: true and consume SSE.
  • Tool use — function-calling round trip.
  • Errors — error envelope shapes.
  • Chat Actions extensions — opt in to Kindo prompts, tools, and stateful conversations.