For AI agents: The documentation index is at https://docs.digitalocean.com/llms.txt. Markdown versions of pages use the same URL with index.html.md in place of the HTML page (for example, append index.html.md to the directory path instead of opening the HTML document).

Action Gateway Tool Playground lets you describe a task to a model and inspect how it searches for and invokes tools. Use it to try a workflow in the Control Panel before integrating it into your application.

Prerequisites

• Access to Action Gateway Tool Playground and a model offered in its model selector.
• Access to any provider accounts the tools need. You can authorize a provider when prompted during the conversation.

Some tools, including Exa web search and web fetch, require a positive prepaid balance. Prepayment is not required for every tool. Model inference has separate billing requirements: DigitalOcean Serverless Inference requires prepayment even when the selected tools do not.

Choose a Model

In the Control Panel, open Managed Agents, select Action Gateway, then open Tool Playground. Select an available model.

The playground creates an Action Gateway session automatically when you send your first message. You do not need to create a session before starting.

Search for and Invoke a Tool

Enter a task in the conversation. For example, ask for discovery without changing resources:

Find a tool that lists available DigitalOcean Droplet sizes. Do not create, update, or delete any resources.

Review the matching tools and their required arguments, then ask the model to run the read-only tool. The model uses Action Gateway to execute it and returns the result in the conversation. Search results can also offer controls for entering arguments and running a tool directly.

If a tool requires a provider connection, follow the connection prompt and sign in to the provider account you want to use. See Authorize Providers for authorization options.

If a tool requires approval, review the tool and arguments before approving it. Authorizing a provider and approving a tool call are separate steps.

Inspect Results and Logs

Review search, invocation, and code result cards in the conversation. Open Logs to inspect model and tool events, arguments, status, and failures. Check each tool’s outcome rather than relying only on the model’s summary.

Open the session and actor panels to inspect the tool access and provider connections used by the conversation.

To use the session from your own application, open the Code view for its MCP connection example. See Connect Action Gateway Tools for client setup.