A focused intelligence layer

Send a prompt.
Get an answer.
Nothing in between.

Ask is a single, model-agnostic API for developers who want intelligence inside their product — without owning a model, a prompt library, or an inference stack to keep running.

POSTask.pinktra.com/v1/ask
POST /v1/askwriting
{ "prompt": " }
Request → response

One call in. One answer out.

Pick a request. Run it. Watch the response come back the same way it would in your app — as it's generated, not after a spinner.

POST /v1/askready
{ "prompt": "Pull the customer name, order ID, and issue from this support message: "Hey, it's Priya — order #48213 arrived damaged, need a replacement."" }
Run a request to see a response stream in.
Built for production

Everything an intelligence layer should be — and nothing it shouldn't.

Single endpoint

One request shape for every kind of prompt. No routing logic, no per-model quirks to learn or maintain.

Model-agnostic by design

Ask selects and manages what runs underneath. Your integration doesn't change when the engine does.

Streaming responses

Read tokens as they're generated, or wait for the full response. Same endpoint, your choice at request time.

Structured output

Ask for a shape, get that shape back — responses your code can parse without a fragile text-matching layer.

Timed, predictable responses

Built for request paths inside real products. Every response returns with the latency it took to produce.

Framework-free

Works with any HTTP client, in any language. No SDK to install and no client library to keep up to date.

Integration

Add intelligence in the time it takes to read this.

No SDK to install. Send JSON to one endpoint, in whatever language your service already speaks.

curl -X POST https://ask.pinktra.com/v1/ask \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ASK_TOKEN" \
  -d '{
    "prompt": "Summarize this ticket in one sentence."
  }'
response.json
{
  "response": "Customer can't log in — password reset link expired before use.",
  "id": "ask_8f2c1d9e",
  "latency_ms": 412
}
How it works

Three steps, one endpoint.

Your appAskResponsepromptresponse
01

Send a prompt

Your application posts a prompt to a single endpoint. No configuration and no model selection required.

02

Ask processes it

Ask interprets the request and handles everything needed to produce the strongest possible response.

03

You get a response

A structured response returns to your app — as text, JSON, or a stream — ready to use immediately.

Where Ask fits

Intelligence, wherever your product needs it.

Ask stays out of the way of what you're building. It's a request your backend makes, not a surface your users see.

In-app assistance

Contextual help and explanations embedded directly in your product's own UI, not a bolted-on widget.

Content pipelines

Generate, rewrite, and summarize copy as a step inside workflows you already run.

Agentic tooling

Give an autonomous workflow a reasoning step, without standing up model infrastructure to support it.

Data extraction

Turn unstructured text — tickets, emails, logs — into structured fields your systems can act on.

Internal tooling

Ship internal automation and admin tools with a reasoning layer, without a dedicated ML team.

Request triage

Classify, prioritize, and route incoming requests automatically, upstream of a human queue.

Developer experience

Documentation that reads like code, not marketing.

Ask is built around one honest contract: a predictable request in, a predictable response out — including when something goes wrong.

  • Consistent JSON on every response, success or failure
  • Typed error codes instead of parsed error strings
  • Sensible request timeouts, returned — not silent
  • No SDK lock-in — call it from anything that speaks HTTP
  • A stable request contract across versions
error.json
{
  "error": {
    "type": "invalid_request",
    "message": "prompt must not be empty"
  }
}
Ready when your app is

Give your product a way to think.

One endpoint. Any prompt. A response your code can use — every time.