-
OrchestrationLangGraph
Graph-based orchestration for stateful, multi-actor agent workflows with explicit control flow and checkpointing.
-
LLM / inferenceLiteLLM
Open-source Python SDK + proxy server (AI gateway) that calls 100+ LLM APIs in OpenAI format with cost tracking, budgets, and virtual keys.
-
MemoryMem0
A memory layer for AI agents — extracts, stores, and retrieves user/agent facts across sessions.
-
Vector storeMilvus
Cloud-native vector database built for billion-scale similarity search.
-
Evals & observabilityLangfuse
Open-source LLM engineering platform — tracing, evals, prompt management, and metrics for agent apps.
Embed this stack
The Open-Source-Only Stack — FAQ
What is Open-Source-Only Stack?
Own every layer — self-host, no vendor lock-in. It pairs LangGraph, LiteLLM, Mem0, Milvus, Langfuse — one tool per job. Teams that want to run the whole stack themselves and bring their own model keys.
Can I change the tools?
Yes — open this stack in the builder and swap any pick, filter by open-source / hosted-API / agent-self-signup, then copy or share the result. Nothing is locked in.
Is it free to use this stack?
The stack, this page, and the builder are free. Each tool has its own pricing, shown on its detail page.
Can an AI agent use this stack?
Yes — it's machine-readable at /api/stack.json?pref=oss, so an agent can request these exact picks (with signup URLs and MCP flags) and act on them.