What happens to agent output when /tmp disappears?

An agent works fine on a laptop, where the filesystem sticks around between runs. Deploy the same agent to Lambda, Cloud Run, ECS, or a GitHub Actions runner, and every invocation starts from a clean container. Anything written to /tmp is gone the moment the function returns, and there's no server to check afterward.

Push the output before the invocation ends, instead of after. One API call, or one line in the SDK, sends the file to Artifacta while the container is still alive. The artifact keeps existing once the container is recycled, because it never depended on that container's disk in the first place.

This is the same push agents use for job-to-job handoffs, so a pipeline that also passes files between agents (see multi-agent file handoff) doesn't need a second mechanism for durability. For agents that publish directly from that same invocation, see scheduled, no-human publishing.

Try it

# inside a Lambda handler, Cloud Run request, or CI job

from artifacta import Client

client = Client()  # reads ARTIFACTA_API_KEY from env

artifact = client.push("/tmp/output.json", metadata={"stage": "final"})

print(artifact.id)  # art_2xk9f7v3m1p0, outlives this container

Live proof

There's no page to click through for this one; it's a durability guarantee at the API layer. The artifact is queryable immediately after the push call returns, independent of whether the container that pushed it still exists.

Get started for free