Tomasz Malisz
Senior Software Engineer, Ingenix.AI
I love this project. Aegra has been a great fit for our stack and the developer experience is excellent.
aegra is an open source Agent Protocol server. Run LangGraph agents on your own Postgres, with the same SDK, and no vendor lock-in.
What you get
Threads, runs, and checkpoints live in a Postgres you control. Query them, back them up, keep them in your region.
Plug in JWT, OAuth, or Firebase with one decorator. Every thread is scoped to its owner.
The endpoints the LangGraph SDK already speaks. Point your client at aegra and keep your code.
Token, value, and event streams over SSE, with resume after disconnect.
A Redis queue with leases, heartbeats, and crash recovery. Scale out by adding instances.
Langfuse, Phoenix, or your own collector. Every run is a trace, every node a span.
Works with the tools you already use
Same SDK
The LangGraph SDK talks to aegra the way it talks to a managed deployment. Change the URL and the client side stays as it is.
from langgraph_sdk import get_client-client = get_client(url=MANAGED_DEPLOYMENT_URL)+client = get_client(url="https://agents.yourcompany.com")thread = await client.threads.create()run = client.runs.stream(thread["thread_id"], "agent", input=state)
Compare
| Feature | LangSmith Deployments | aegra |
|---|---|---|
| Deploy agents | Local dev only (Free), paid cloud (Plus) | Free, unlimited |
| Custom auth | Not available (Free), available (Plus) | Python handlers: JWT, OAuth, Firebase |
| Self-hosted | Enterprise only, license key required | Always, Apache 2.0 |
| Own database | Managed only (Free/Plus), BYO (Enterprise) | Bring your own Postgres |
| Tracing | LangSmith built in, others via callbacks in your code | Any OTLP backend as a server setting: Langfuse, Phoenix, more |
| Data residency | LangChain cloud (Free/Plus), yours (Enterprise) | Your infrastructure |
| SDK | LangGraph SDK | Same LangGraph SDK |
Based on LangChain pricing as of February 2026.
Tomasz Malisz
Senior Software Engineer, Ingenix.AI
I love this project. Aegra has been a great fit for our stack and the developer experience is excellent.
Soufiene Slimi
Discord
Aegra really nails that gap. A self-hosted, open source LangGraph Platform alternative with zero vendor lock-in, your own Postgres, and the same LangGraph SDK is exactly what I was hoping someone would build.
Geoffrey
Diag n'Grow
We use Aegra for our agents at Diag n'Grow: a data collection chatbot that replaces a fixed questionnaire, and a software audit agent.
Artur Bagramyan
Head of Agentic AI Engineering, ROQ Capital
At ROQ Capital, we use Aegra in production as a core component of the agentic AI infrastructure behind ROQ Markets, our AI-powered investment platform. Aegra provides the flexibility, transparency, and control we need to build and scale complex AI systems while leveraging the LangGraph ecosystem.
The whole server is on GitHub. Read it, run it, fork it. There is no enterprise tier hiding the good parts, and no license key. If you can run Postgres, you can run aegra.
View on GitHub