Use case

Human-in-the-loop review

Some decisions need a person. Pause any node, let a human inspect and edit state, then resume exactly where you left off.

In one line: HITL, interrupts and time travel are part of the core StateGraph surface — no platform tier required. Studio is a separate open-source package that mounts on any running agent.

The problem

High-stakes agent actions need review, but most runtimes make pausing and resuming awkward.
Once an agent has run, it's hard to see why it chose a branch — or to correct it.
Approval steps often mean rebuilding state by hand outside the framework.

How Agentmatic solves it

Interrupt before or after any node

interrupt_before and interrupt_after pause execution at a boundary so a reviewer can approve, reject, or edit the state before the graph continues — the same API as LangGraph.

Edit state and resume

Because state is checkpointed at every superstep, a human can change it in place and resume, or rewind with time travel to try a different path.

See it in Studio

Agentmatic Studio, the open-source visual debugger, gives you a step debugger, checkpoint timeline, and state-diff viewer over a running agent.

Keep reading

Part of the Agentmatic use cases, from Neul Labs ↗.

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.

Migrate from LangGraph → Star on GitHub ★