Yaha AI
Workflows

Workflows Overview

Visual DAG automations that call Yaha Agents, APIs, and LLM steps on a governed platform.

A Yaha Workflow is a versioned directed graph you build in the console under Products → Workflows. One trigger starts a run. Nodes transform data, call HTTP APIs, complete chat requests through the Gateway, or invoke a Yaha Agent. An output node returns the result.

Drafts are editable. Publish pins an immutable version so later runs stay deterministic.

When to use a workflow

  • Chain several steps that a single agent prompt cannot own reliably
  • Mix integrations (HTTP, chat platforms) with LLM and agent steps
  • Run the same process on a schedule or from a webhook
  • Require a human to approve before the flow continues

For a single assistant you chat with or embed, start with Yaha Agents. For Cursor or Claude Code, use coding agent integrations.

Console surfaces

PagePurpose
Products → WorkflowsList, create, and open the canvas
Workflow editor /products/workflows/{id}Build the graph, save draft, publish, run
RunsHistory, status, tokens, and cost
ApprovalsReview gates that paused a run

How workflows use agents

Add an Agent invocation node and select a published agent. The workflow passes input into that agent and continues with the agent output. Chat-platform templates can use a WhatsApp-style agent node that keeps a session per sender.

Agents and workflows share the same workspace, Gateway models, and usage tracking.

Next steps

On this page