Product

Understand.
Execute.

Setu is an API-first workflow layer. It reads a message, resolves a workflow, calls your APIs, and returns a verified business result.

01 · UNDERSTAND

Natural language, in.
Intent + entities, out.

The Setu API accepts the raw message plus optional session context. It returns a typed intent, structured entities, and a confidence score. Your app decides what to do next.

  • Schema-locked structured output. Predictable, versioned.
  • Multilingual examples include English, Hindi, Hinglish, and languages with strong expected performance.
request
Hinglish example

"return karna hai, order FL1234567890, size chhoti aa gayi"

Setu API · response
Hinglish example
{
  "intent": "return_request",
  "entities": {
    "order_id": "FL1234567890",
    "reason": "size_issue"
  },
  "language": "hinglish",
  "confidence": 0.94
}
02 · EXECUTE

One request.
Whole workflow run.

The Setu API takes the same input, but runs the full workflow: input validation, calls to your APIs, conditional branching, and a structured business response.

  • Deterministic step execution. AI never improvises past the workflow.
  • Semantic actions map to your existing endpoints.
  • Guardrails and confirmations for high-impact actions.
Setu API · trace
{
  "request_id": "req_9f4e",
  "intent": "return_request",
  "workflow": "return_pickup",
  "actions_taken": [
    "get_order",
    "check_return_eligibility",
    "create_return_pickup"
  ],
  "outcome": "pickup_scheduled",
  "pickup_window": "2-3 days",
  "refund_days": "2-3 working"
}
optional reply
Hinglish example

Return request register kar li hai. Pickup 2-3 din mein schedule ho jayega, aur refund 2-3 working days mein aapke UPI par aa jayega.

FAQs

AI workflow API questions

What does Setu’s AI workflow API do?

Setu interprets a customer message, extracts the information a declared workflow needs, validates inputs, and can run that workflow through your existing APIs. It returns structured results your application can use.

Does Setu replace our backend or business logic?

No. Your CRM, ERP, commerce, and other systems remain the source of truth. Setu maps workflow actions to endpoints in the systems you already operate.

How does Setu keep AI actions controlled?

You declare the workflow steps, conditions, and required inputs. Setu validates the request against that workflow, and higher-impact actions can use permissions or human confirmation.

Which languages can Setu understand?

The current product experience describes support for English, Hindi, and Hinglish.

Two ways to build

Chatbot theater,
or verified execution.

Most AI stacks today are prompts glued to hope. Setu replaces that with deterministic workflows, traced actions, and versioned intents, everything you'd already demand from the rest of your infrastructure.

AWithout Setu

Chatbot glued to hope.

  • UnderstandingPrompts + guesswork. Every model change is a rewrite.
  • ExecutionAd-hoc tool calls. AI may or may not do the right thing.
  • Accountability"AI did it." No trace, no audit, no request_id.
  • New intentsWeeks of prompt engineering and manual QA.
  • DebuggingGrep through chat logs, rerun by hand, cross fingers.
  • CostPer-seat SaaS or per-token chaos. Both scale badly.
BWith Setu

Verified business execution.

  • UnderstandingSchema-locked intent + entities on every request.
  • ExecutionDeterministic workflow steps mapped to your APIs.
  • AccountabilityEvery request has a request_id. Every action logged.
  • New intentsOne-click merge from the adaptive queue.
  • DebuggingEnd-to-end trace: request → intent → workflow → API → result.
  • CostPriced against traffic and deployment mode, not seats.
What's next

The API keeps growing.

Sessions, webhooks, programmatic workflow management, and multi-region residency. All designed to keep the integration contract stable.