An approval pauses the run, not a workerA paused run releases its lease and holds no worker. It waits until someone decides, then the approval node returns the decision and the run continues from that step.Why every execution is pinned to one versionAn execution records its workflow version once, when it starts, and never follows a live pointer. That is why resume and replay still mean something weeks after a graph has changed.Memory has a subject, a namespace, and an expiryEvery memory fact lives under a key, a subject and a namespace, and it always expires: 90 days by default, one year at most. Here is how writes, reads and sharing work.What happens when a worker dies mid-runA worker holds a 30 second lease and renews it every 10 seconds. If it dies, the lease expires, another worker claims the run, and the run resumes from the last checkpoint.Which Linea client to use, and which key goes whereThe TypeScript SDK has a server client, an environment client for production starts, an end-user client for browsers, and a webhook verifier. Each one takes a different credential.Linea Labs is not Consensys LineaLinea Labs builds Linea, an open-source AI agent runtime at getlinea.app. It is not the Consensys Linea blockchain, and not the Linea Sketch iPad app.What is an AI agent runtime?A framework gives you the agent loop. A runtime runs it in production: durable steps, checkpoints, approvals, crash recovery, and replaying one failed step.Why replay needs the checkpointYou cannot re-run step four without the exact input it saw. That input only exists if the host executed the run and kept the checkpoint.One graph, not two productsThe canvas, the workflow JSON, and the node registry describe one thing. A second editor is fine. A second graph is a second product.The half frameworks do not shipFrameworks solve the loop. You still host it, retry it, store its memory, and keep its checkpoints. That second half is the product.Memory is not a side questFacts have to survive the run. If you stand up the store yourself, you are still doing the ops half. We think the graph should ask for memory and the host should keep it.Sandbox the code nodeA code node is an escape hatch. It still has to run somewhere that cannot take down the host. Linea does not ship one yet, and this is why.Triggers are not the productStarting a graph is a solved problem. The product is what has to stay up after it starts: the worker, the checkpoint, the wait, and the memory.
Toggle menu

Architecture

The host is the graph, the worker, and the checkpoint.

Linea executes the version you published. It does not observe a loop running somewhere else, so the checkpoint belongs to the host.

A Linea workflow graph on the canvas
published version: the canvas, the checkpoint and the replay target share its node ids
1 graphpublished version: the canvas, the checkpoint and the replay target share its node ids
worker lease: if a worker dies, another picks the run up
30 sworker lease: if a worker dies, another picks the run up
records its input and output when it completes
Every noderecords its input and output when it completes
one data model: graph, API, workers and checkpoint
4 piecesone data model: graph, API, workers and checkpoint

What runs

Four pieces, one data model.

The same node ids are on the canvas, in the checkpoint, and in the replay target.

  1. 01

    One published graph

    The builder saves the workflow as a document. Publish freezes a version. Checkpoints, replay, and regression cases all use the node ids in that version.

  2. 02

    An API that starts the run

    A manual run, a webhook, a schedule, or the TypeScript SDK starts one execution of the published version. Starting the graph is not the product. The run after that is.

  3. 03

    Workers on a 30-second lease

    Queue-backed workers hold a lease on the run they are executing. If a worker dies, the lease expires and another worker picks the run up.

  4. 04

    A checkpoint per completed node

    Each finished node records its input and output. Resume continues from the last successful step. Replay loads that same input for one node.

A crash resumes the run

When the worker holding the lease dies, another worker starts from the last checkpoint. Completed steps stay completed. The step that was in flight runs again.

A wait holds the checkpoint

An approval or a wait pauses the execution at the checkpoint until someone decides or a timer fires. The run then continues from that step. It does not start over.

Questions

Questions about this page.

How is this different from LangGraph?

A framework helps you write the loop. You still host the process, store checkpoints, keep memory, and operate the result. Linea is that host and control plane. Keep the prompts and tools you already have; we execute the graph.

What is Linea?

Linea is an agentic runtime and control plane. You ship an agent to your users. Linea executes the workflow graph behind it — model calls, tools, waits, approvals, and memory — and keeps the records around the run: checkpoints, signals, approvals, and regressions.

You write the agent.
We execute the graph.

The Linea newsletter

What we learn from every run.

New releases and field notes on running agents in production.

One email when something ships. Unsubscribe anytime.