Action Catalog
The Action Catalog is the library of third-party API operations that ToolSmith searches while it researches and builds HTTP Actions. For each operation, the catalog holds the documentation ToolSmith needs to configure an action: the operation itself, its parameters, field definitions, and request and response schemas.
You describe what you need in plain language, for example “look up a Jira issue by key” or “create a Zendesk ticket”. ToolSmith finds the matching operation in the catalog and uses its documentation to fill in the HTTP action. You still set up the connector and credentials, decide who can use the action, and test it before it goes into a plugin.
Why it matters
Without the catalog, building an HTTP action starts with research: find the right endpoint in the vendor’s documentation, work out the parameters, and shape the request and response by hand. With the catalog, ToolSmith already has that documentation for the operation you describe, so:
- The research step is skipped. ToolSmith goes from your description to a candidate operation without you opening vendor docs.
- The request shape is right the first time. Parameters, field definitions and response schemas come from the published specification, not from guesswork.
- Test cycles go to your data, not to the API. You spend iterations on permissions, inputs and the response you want, rather than on discovering how the API works.
The result is a working HTTP action in far fewer steps, and one that matches the vendor’s documented contract.
What the catalog contains
The catalog covers 262 systems and more than 60,000 documented actions. The documentation comes from each vendor’s published API specifications, including OpenAPI and Swagger, Google Discovery, and JSON Hyper-Schema documents.
Vendors that publish more than one API are grouped under a single system. Jira, for example, covers both the Jira Platform and Jira Software APIs, and Microsoft covers 20 API families under one entry.
A few systems in the catalog are narrower than their name suggests:
- Buildkite covers the Test Engine API only, not the main Buildkite REST API.
- Firebase covers the Firebase Management API only, not everything in the Admin SDK.
- Mailgun covers the InboxReady API only, not the Mailgun Email API.
If you need an operation outside these surfaces, you can still build the HTTP action manually. See What if my system isn’t covered?
What being in the catalog means
A system in the catalog means ToolSmith has that system’s API documentation and can use it to draft the request, parameters, and response handling for an HTTP action.
Catalog coverage does not mean every operation has been executed against a live instance. ToolSmith builds from the vendor’s published documentation, so always test the action in Agent Studio with your own connector before adding it to a plugin. See HTTP Action Troubleshooting if a test fails.
The catalog also does not replace the rest of HTTP action setup. You still need to:
- Create or select a connector with the base URL and credentials for your instance.
- Configure permissions and access control for the plugin that uses the action.
- Test the action with real inputs and check the response.
How to check whether a system is covered
Open the AI Agent Marketplace Explore page and apply the ToolSmith filter. Every system in the catalog appears in the results, and each system’s page lists the actions ToolSmith can build for it.
Check the system’s page before you start. If the operation you need is listed, ToolSmith can build it from the catalog. If it is not listed, you can still build the HTTP action manually.
FAQ
What is the Action Catalog?
A library of third-party API operations and their documentation that ToolSmith searches while building HTTP actions. It covers 262 systems and more than 60,000 actions, drawn from published API specifications.
How do I know if my system is covered?
Go to the AI Agent Marketplace Explore page and filter by ToolSmith. Open the system’s page to see the actions ToolSmith can build.
Do I still need a connector?
Yes. The catalog provides API documentation only. Your connector supplies the base URL and credentials for your instance, and every HTTP action ToolSmith builds is attached to a connector you configure. See Connectors.
What if my system isn’t covered?
You can build the HTTP action yourself in Agent Studio by following HTTP Actions, either by importing a cURL command from the vendor’s documentation or by configuring the request manually. You can also request coverage for the system from the AI Agent Marketplace.