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

SDK

Start the run from your app.

The SDK is TypeScript. It starts executions and reads them back, and a server key can create workflows and save drafts and versions. Publishing needs a signed-in admin in the dashboard, and the host runs the version that was published.

Linea dashboard listing executions started from the app
the only SDK: there is no Python client
TypeScriptthe only SDK: there is no Python client
what it does: starts executions and reads each step's record back
Start + readwhat it does: starts executions and reads each step's record back
the SDK can save drafts and versions, but publishing and binding an environment need a signed-in admin
Admin publishesthe SDK can save drafts and versions, but publishing and binding an environment need a signed-in admin
verified by the SDK, alongside starting and reading runs
Webhooksverified by the SDK, alongside starting and reading runs

The handoff

Three things cross from your app to the host.

  1. 01

    Name the version

    When you publish, the builder freezes the workflow as a version: the graph the host will run. Your app names that version and asks the host to start one execution of it, so the graph you reviewed is the graph that runs.

  2. 02

    Pass the input

    The start carries the input the trigger expects: a webhook body, the fields the first node reads, or the id of the person the run is for. It is data for the graph, not a description of it. Prompts, tools, and edges stay on the host.

  3. 03

    Read the record

    As the run proceeds, the host writes each step's input and output onto the execution. The SDK reads that record back: the same record you see in the run, and the same checkpoint replay loads for one node.

One graph

Your app starts the run. An admin publishes the graph.

If the nodes in a run are not the nodes on the canvas, replay has nothing to target. The SDK can prepare a draft or a version, but only an admin publishes one, and the host runs that published version.

01

What the SDK does

Start an execution and read each step's record back. A server key can also save drafts and versions.

02

What an admin does

Publishing a version and binding an environment need a signed-in admin. The host runs that version.

03

TypeScript only

There is no Python client. A payload sample in another language is an example of the body, not an SDK.

The Linea builder canvas, where the graph is authored

Questions

Questions about this page.

Do I have to use the visual builder?

No. Every published version is a JSON graph the host runs exactly as published. You can author that graph in the visual builder, or create workflows and save drafts and versions through the TypeScript SDK with a workspace API key. Publishing a version and binding it to an environment need a signed-in admin in the dashboard. The SDK also starts executions, reads their results, and verifies webhooks.

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.

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.