Studio
See your graph think.
A visual debugger that mounts on top of any running Agentmatic agent. Step through supersteps, inspect state diffs, branch from any checkpoint. Open-source, self-hosted.
What it does
- Graph visualization — nodes, edges, conditional routes rendered as a force-directed diagram.
- Step debugger — pause before any node; inspect inputs, outputs, and state diffs; resume.
- Checkpoint timeline — scrubbable history of every superstep, with branch-from-here.
- Tool call inspector — every LLM call, every tool invocation, with prompts and raw responses.
- State diff viewer — diff between any two checkpoints, JSON-friendly.
Run it
pip install agentmatic-studio
agentmatic-studio serve # http://localhost:7860 Then attach your agent:
from agentmatic.studio import attach
agent = Agent.builder("debug").llm(OpenAI()).tools([...]).build()
attach(agent, port=7860) # opens Studio in browser
await agent.ainvoke(...) Compared to LangGraph Studio
- Open source. Apache-2.0, self-hosted, no SaaS dependency.
- Local-first. All state stays on your machine.
- Multi-language. Attaches to Python, TS, Rust, Go, and Java agents over the same gRPC interface.
Ship your next agent in minutes, not weeks.
MIT licensed. Drop-in for LangGraph. Native SDKs in 5 languages. Battle-tested resilience primitives in the box.