Build Index / Open Source scaffold

Build Index by Enterprise DNA

Open Source

The pieces engineers reach for when "buy" is not the answer. Orchestration graphs, validation libraries, eval harnesses, observability. Tracked here with stars, recency, and what real teams ship on top of them.

Entries indexed

OSS

AutoGen

Microsoft

Microsoft's framework for multi-agent conversations. Agents that talk to each other to solve hard problems.

open-source

Best for: Teams comfortable with multi-agent patterns, on Python or .NET

Framework medium

BAML

Boundary

A typed language for LLM functions. Define inputs, outputs, and prompts, get reliable structured output.

open-source

Best for: Teams who want LLM calls to feel like real typed functions

Framework medium

CrewAI

Role-based multi-agent framework. Define crews of agents with roles, goals, and tasks, run them as a team.

open-source

Best for: Teams who think in roles, not graphs

Framework high

DSPy

Stanford NLP

Programming, not prompting. Declare what you want, compile prompts and weights against an objective.

open-source

Best for: Teams who want to optimise their LLM pipelines like code, not edit prompts forever

Orchestration low

Inngest

Durable workflows for agents. Step functions, event-driven runs, retries, all without managing a queue.

freemium

Best for: Teams whose agents need to be durable, not just clever

Framework one click

Instructor

Jason Liu (community)

Structured output for LLMs via Pydantic. The cleanest answer to 'just give me a typed object back'.

open-source

Best for: Engineers tired of regexing JSON out of model output

Obs low

Langfuse

Open-source LLM observability. Traces, evals, prompt management, all self-hostable.

freemium

Best for: Teams running LLM apps in production who need real observability

Orchestration medium

LangGraph

LangChain

Graph-based orchestration for long-running, multi-step agents. The control plane LangChain always needed.

open-source

Best for: Engineering teams building production agent systems, not weekend prototypes

Memory medium

Letta

Stateful agents with first-class memory. The continuation of MemGPT, productionised.

open-source

Best for: Anyone building persistent agents whose memory matters

RAG medium

LlamaIndex

The data framework for LLM apps. RAG, ingestion, structured extraction, agents over your data.

open-source

Best for: Teams whose agent value comes from their own data, not just the model

Framework low

Mastra

TypeScript-first agent framework. Workflows, agents, tools, memory, evals, in one consistent shape.

open-source

Best for: TypeScript teams who want a single agent framework, not a stack

Memory low

mem0

Memory layer for AI apps. Personalisation, continuity, and recall as a service.

freemium

Best for: Product teams who want memory without building the storage layer

Framework medium

Semantic Kernel

Microsoft

Microsoft's enterprise-flavoured framework for AI agents. .NET-first, with Python and Java siblings.

open-source

Best for: .NET and enterprise teams adopting AI agents

Orchestration high

Temporal

Durable execution at enterprise scale. The control plane that survives anything.

freemium

Best for: Enterprises and platform teams running serious multi-agent workloads

Orchestration low

Trigger.dev

Background jobs and long-running tasks for AI apps. Write functions, get durability, observability, and retries.

freemium

Best for: TypeScript teams shipping long-running agent jobs in production

SDK one click

Vercel AI SDK

Vercel

The de facto TypeScript SDK for AI apps. Streaming, tools, multi-model, and now an agent loop.

open-source

Best for: TypeScript teams shipping AI features in production