Skip to content

LangGraph server · embeddable · Apache-2.0

Ship your LangGraph as a real HTTP API.

skeino is a reusable, embeddable replacement for the langgraph dev server — threads, runs, streaming, and assistants — the LangGraph Studio surface over any graph you own.

Get started How it works GitHub ↗
  • FastAPI
  • SSE streaming
  • Pluggable checkpointers
  • Python 3.11+
app.py 
## Your graph, served over the LangGraph HTTP API
from skeino import create_app, SkeinoSettings
from my_project.graph import graph

app = create_app(
    graphs={"my_agent": graph},
    settings=SkeinoSettings(
        postgres_uri="postgresql://localhost/skeino",
    ),
)
$ uvicorn app:app --port 8000

skeino

Why skeino

  • Studio-compatible

    Implements the v1 HTTP surface — threads, runs, streaming/SSE, assistants, health/info — that LangGraph SDK clients and Studio already speak.

  • Modular by design

    api, ops, persistence, streaming, and serialization are separate concerns with explicit dependencies — easy to read, easy to extend.

  • Pluggable persistence

    Checkpointers register through a small decorator-based registry. Postgres and in-memory ship in the box; add Redis or Mongo without touching core.

  • Typed & documented

    Strict mypy and enforced docstrings across the package — so the Python API reference comes straight from source.

Two ways in

from skeino import create_app, SkeinoSettings
from my_project.graph import graph

app = create_app(
    graphs={"my_agent": graph},
    settings=SkeinoSettings(
        checkpointer_scheme="postgres",
        checkpointer_uri="postgresql://localhost/skeino",
    ),
)

from skeino import from_langgraph_json

app = from_langgraph_json("langgraph.json")

Then serve it like any ASGI app:

uvicorn app:app --port 8000

Your graph is now reachable over the LangGraph HTTP API, with interactive OpenAPI docs at /docs — and an always-current API explorer right here in these docs.

Explore the docs

  • Getting started

    Install skeino and stand up a server — programmatically or from langgraph.json.

  • Concepts

    The thread/run/checkpoint model, streaming semantics, and persistence.

  • API reference

    Every v1 HTTP endpoint, the interactive explorer, and the generated Python API.

  • How-to guides

    Embed in an existing app, configure Postgres, write a checkpointer, deploy.

Status

skeino is stable (1.x, semver). The supported public surface is create_app, from_langgraph_json, SkeinoSettings, and GraphRegistry. Sub-modules are importable for advanced use but aren't part of the stability contract. See the changelog for release history.

Some langgraph dev features are intentionally out of scope for v1: no bundled web UI, no cron/webhook scheduling, no built-in auth, and no /store/* endpoints. skeino focuses on the core threads/runs/streaming/assistants surface that SDK clients depend on.