Client

Kitaru is ZenML's sibling project for running, replaying, and improving AI agents. KitaruClient is the programmatic interface to a Kitaru server — inspect and control flow executions, browse artifacts, invoke deployments, and manage authentication from Python.

This page groups the client into thematic areas. For the complete, auto-generated API (every method and signature), see the full Kitaru SDK reference.

import kitaru

client = kitaru.KitaruClient()

The client exposes its functionality through sub-APIs accessed as attributes.

Executions

client.executions inspects and controls flow executions:

• get — fetch a single execution by ID
• list — list executions
• latest — the most recent execution
• logs — retrieve runtime logs for an execution
• input — provide input to an execution waiting on kitaru.wait()
• resume — resume a waiting execution after input lands
• retry — retry a failed execution
• cancel — cancel a running execution
• replay — replay from a checkpoint, optionally overriding cached outputs

Artifacts

client.artifacts browses and reads checkpoint outputs and saved artifacts.

Deployments

client.deployments invokes and inspects deployed flows:

• invoke — start a new execution from a deployed flow (by name + tag/version)
• list / get — inspect available deployment routes

Authentication

client.auth manages service accounts and API keys for server access.

---

For complete signatures and the full method list, see the Kitaru SDK reference. For the command-line equivalents, see the Kitaru CLI reference.