Official SDKs

Integrate Enprompta into your applications with our official SDKs. Choose the one that fits your stack.

TypeScript SDK

@enprompta/sdk

Auto-instrument OpenAI, Anthropic & Gemini, serve versioned prompts, and ship traces — with middleware, retries, and full type safety.

Auto-InstrumentationType SafetyMiddleware SystemOAuth2 Support

npm install @enprompta/sdk

Python SDK

enprompta

View Docs

Auto-instrument OpenAI, Anthropic & Gemini in one call, plus an async-first client built on httpx with pydantic models.

Auto-InstrumentationAsync/AwaitPydantic ModelsSync Wrapper

pip install enprompta

Quick Comparison

Feature	TypeScript	Python
API Key Auth	Yes	Yes
OAuth2 Auth	Yes	Yes
Automatic Retry	Yes (6 strategies)	Yes
Middleware System	Yes (extensible)	Yes (basic)
Type Safety	Full TypeScript	Type hints + Pydantic
Async Support	Native async/await	Async + Sync wrapper
Auto-pagination	Async generators	Async generators
Minimum Runtime	Node.js 18+	Python 3.8+

Quick Start

TypeScriptFull docs →

import { Enprompta } from '@enprompta/sdk'

const client = new Enprompta({
  apiKey: process.env.ENPROMPTA_API_KEY
})

const prompts = await client.prompts.list()

const result = await client.prompts.execute(
  'prompt_id',
  {
    variables: { name: 'World' },
    provider: 'openai',
    model: 'gpt-4'
  }
)

PythonFull docs →

from enprompta import Enprompta

client = Enprompta(
    api_key=os.environ["ENPROMPTA_API_KEY"]
)

prompts = await client.prompts.list()

result = await client.prompts.execute(
    "prompt_id",
    variables={"name": "World"},
    provider="openai",
    model="gpt-4"
)

Text vs structured prompts

A prompt can be authored two ways. Most start as text and move to structured when they need a separate system instruction or distinct roles. Modern chat models treat the system role specially, so keeping it separate from the user template produces better, more predictable behaviour than flattening everything into one message.

Type	Authored as	Served as	Use when
Text (Simple)	One prompt body	`content` string	A single instruction or template
Structured	System + User messages	`messages` (roles) + composed `content`	Separate system instructions or roles

The runtime fetch always returns content (for a structured prompt it's the System and User text composed together, so existing integrations keep working) and messages (the structured form, or null for a text prompt). Prefer messages when present:

const p = await client.prompts.getLive('summariser', { label: 'prod' })

// Works for both prompt types:
const messages = p.messages ?? [{ role: 'user', content: p.content }]
await openai.chat.completions.create({ model: 'gpt-4o', messages })