Deployments
Publish a Yaha Agent, then deploy it to API, widgets, chat platforms, MCP, or workflows.
Deployments attach a published agent version to a channel. Create the agent and publish a version first, then open the builder Deployment section.
Channels
The console offers these channels:
| Channel | Typical use |
|---|---|
| REST API | Call the agent from your app with a deployment key |
| Webhook | Receive events and invoke the agent |
| Website widget | Embed a chat widget on a site |
| Slack, Teams, Discord | Chat-platform bots |
| MCP server | Expose the agent as an MCP server |
| Workflow | Bind the agent so a workflow can invoke it |
Available channels can depend on your plan. The console only lists channels your workspace can use.
Create a deployment
- Confirm the agent has a published version.
- Open Deployment on the agent.
- Choose a channel and environment (for example, production).
- Deploy. For the REST API channel, the console shows a deployment API key once.
Store that key like any other secret. You can rotate it later from the same section. Do not commit deployment keys to source control.
Call a deployed agent (REST)
External callers use the deployment key on the public agent chat endpoint. The console deployment panel shows the exact URL and header for your environment.
Typical pattern:
curl https://app.yahagateway.io/public/agents/chat \
-H "Authorization: Bearer YOUR_DEPLOYMENT_KEY" \
-H "Content-Type: application/json" \
-d '{"input": {"text": "Hello from the deployed agent"}}'Use the URL and input shape shown in the console if they differ for your workspace. Project yaha.* Gateway keys are for the OpenAI-compatible Gateway, not for this deployment chat endpoint.
After deploy
- Runs still appear under Products → Agents → Runs.
- Update the agent by editing a draft and publishing a new version, then point the deployment at that version if the console asks you to.
- Workflows can invoke the same published agent through an Agent invocation node without a separate REST client.