Hero

Ship Production AI
in Minutes, Not Months

The TypeScript framework for building, deploying, and managing AI agent systems
$npm install orkajsFULL
or
$npm install @orka-js/core @orka-js/*SELECTIVE

+20 modular packages — install only what you need

quick-start.ts

Compatible Runtimes & Frameworks

Express
Express
NestJS
NestJS
Hono
Hono
KoaJS
KoaJS
Fastify
Fastify
AdonisJS
AdonisJS
SailsJS
SailsJS
MeteorJS
MeteorJS
LoopBack
LoopBack
Express
Express
NestJS
NestJS
Hono
Hono
KoaJS
KoaJS
Fastify
Fastify
AdonisJS
AdonisJS
SailsJS
SailsJS
MeteorJS
MeteorJS
LoopBack
LoopBack

OrkaJS is framework-agnostic — use it with any Node.js or Bun runtime

Why developers choose Orka.JS ?

Built for quick and easy setup,
no complications.

Intent-Based API

orka.ask(), orka.agent(), orka.workflow() — code that reads like English.

Swap Provider

Switch models with one line of code.

Production-Ready

Retry, fallback, and observability built-in.

Fully Typed

End-to-end TypeScript with full IDE autocomplete.

Evaluation

CI/CD test runners to prevent regressions.

Everything you need
to ship AI.

From ingestion to evaluation — the full lifecycle of production LLM applications in one TypeScript-first framework.

Explore Documentation

Model Context Protocol

Connect your AI to external tools and data sources with standardized MCP servers.

const mcp = orka.mcp({
servers: ['filesystem', 'github'],
tools: ['read_file', 'search_code']
});

Fine-Tuning

Train custom models on your data with built-in dataset preparation and evaluation.

const job = await orka.fineTune({
model: 'gpt-4o-mini',
dataset: trainingData,
epochs: 3
});

RAG & Knowledge

Connect your data with high-level abstractions for embeddings and vector stores.

const knowledge = await orka.knowledge({
vectorStore: 'pinecone',
index: 'docs'
});

Autonomous Agents

Define tools and let Orka handle the planning, execution, and memory management.

const agent = orka.agent({
model: 'gpt-4o',
tools: [search, database]
});

Multimodal AI

Process images, audio, and text together with Vision and Audio agents.

const visionAgent = new VisionAgent({
llm,
systemPrompt: 'Analyze images'
});
await visionAgent.analyzeImage(image);

Human-in-the-Loop

Require human approval for sensitive operations with checkpoints and resume.

const agent = new HITLAgent({
requiresApproval: ['transfer_money'],
onInterrupt: approvalHandler
});

Real-time Streaming

Stream tokens as they're generated for responsive AI applications.

await llm.streamGenerate(prompt, {
onToken: (token) => {
process.stdout.write(token);
}
});

PII Protection

GDPR-compliant data protection layer to detect and redact sensitive information.

const guard = new PIIGuard();
const safe = guard.redact(
'Email: user@example.com'
);
// → 'Email: [EMAIL]'

Production Evaluation

Built-in assertions and test runners to quantify your LLM performance.

await orka.evaluate(result, {
metrics: ['faithfulness', 'latency'],
threshold: 0.8
});

Multi-Model Workflows

Chain models, handle fallbacks, and orchestrate complex AI logic seamlessly.

const flow = orka.workflow()
.step('summary', modelA)
.step('critique', modelB)
.execute();
Runtime Orchestration

One interface,
every LLM provider.

Stop worrying about rate limits or provider downtime. Orka handles the logic of multi-model resilience at the core level.

RouterLLM

Complexity-aware routing

RaceLLM

Lowest latency wins

ConsensusLLM

Best-of-N Majority Vote

FallbackLLM

Auto-healing failover

orchestration.ts
Zero to Production

Ready to build
the future of AI?

Join the developers building autonomous, resilient, and type-safe LLM applications with Orka JS.

npm install orkajs|npm install @orka-js/*
v3.4.5 — 21 modular packages