Choose how to author an automation

Change an automation's fields directly in the editor, or let your coding agent build and change it through Tale’s MCP endpoint. Tale has no AI assistant inside the editor: a coding agent such as Claude Code or Codex, connected with your API key, is how you work on automations with AI. Both routes save versions of the same workflow and use the same validation and deployment rules. Authoring and deploying automations take the Owner, Admin or Developer role.

Make a change in the editor

Open Automations and select the workflow. Select a node to inspect its input, model, code, or other configuration, and change the field you need. Save the change with a version message, run a test, and deploy the intended version when its checks pass. The canvas arranges itself from the references between nodes; you don't add nodes or draw connections on it.

The workflow editor covers these steps in detail, including reading the canvas, inspecting a run and returning to an earlier version. The canvas does not include a conversational assistant panel.

Change it with your coding agent

Edit with your coding agent, the last button at the top right of the editor's canvas, is the way in. Its dialog shows the automation's name to give the agent, Set up MCP, which opens Settings > API > MCP, and a link to the guide for connecting a coding agent. Connect your coding agent with the endpoint under Settings > API > MCP and your personal API key; Use Tale from your editor or a script has ready configurations. Tell it what the automation should take in, what it should produce, and which systems it may change. Ask it to look at the existing automations and at what your organization has before it creates another one.

Your agent works on an automation the way you do in the editor. It reads the reference and the current version, validates its change, runs it on the mocks and runs the automation’s tests, then saves a new version that names the version it started from. If someone saved a newer version in the meantime, Tale refuses the save, and the agent reads that version and merges its change first. When it starts a saved version on the mocks, the run appears in the automation’s Runs tab as a Test run, Started by you (API), so you can open what it ran. A version the agent saves appears on the canvas while you look at the automation.

Saving creates a version; it does not make it live. Before your agent deploys, deletes, sets a trigger or installs an automation in projects, a client that honors Tale’s marking for these tools, such as Claude Code, asks for your yes, even when you let the agent run other tools without asking. Review the version and its test results before you agree. Every change your agent makes is in the audit log with Source Coding agent, and the MCP endpoint lists every tool it can use.

Keep runtime decisions separate

An agent node inside an automation performs work during a run. It is separate from the coding agent you use to write the workflow. Similarly, an approval authorizes one pending operation during execution; it does not approve a proposed edit to the workflow definition. Your coding agent never decides an approval: a run it starts that needs one waits until a person decides in Tale.

Choose the editor for a change you want to make directly, or MCP for authoring with your coding agent. Start from an existing automation when a suitable one is already available.