> This page is for API Reference & Tools.

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.moveworks.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.moveworks.com/_mcp/server.

# Tool Studio

> Build Tool Studio connectors, actions, data types, conversational processes, plugins, webhook listeners and MCP servers through the Configuration API.

The Tool Studio entities live under `/platform/extensions`. They are the same objects you build in Tool Studio, so you can create and update them from code instead of the Tool Studio UI.

| Entity                                                                                                     | Path                                            | What it is                                         |
| ---------------------------------------------------------------------------------------------------------- | ----------------------------------------------- | -------------------------------------------------- |
| [HTTP Connectors](/api-reference/configuration-api/tool-studio/http-connectors/overview)                   | `/platform/extensions/http-connectors`          | Base URL and credentials for an external API       |
| [HTTP Actions](/api-reference/configuration-api/tool-studio/http-actions/overview)                         | `/platform/extensions/http-actions`             | One request to that API                            |
| [Scripts](/api-reference/configuration-api/tool-studio/scripts/overview)                                   | `/platform/extensions/scripts`                  | Python code run as an action                       |
| [Compound Actions](/api-reference/configuration-api/tool-studio/compound-actions/overview)                 | `/platform/extensions/compound-actions`         | Several steps run without the user                 |
| [Data Types](/api-reference/configuration-api/tool-studio/data-types/overview)                             | `/platform/extensions/data-types`               | A business object and how to look it up            |
| [Conversational Processes](/api-reference/configuration-api/tool-studio/conversational-processes/overview) | `/platform/extensions/conversational-processes` | The user-facing flow: slots and steps              |
| [Plugins](/api-reference/configuration-api/tool-studio/plugins/overview)                                   | `/platform/extensions/plugins`                  | What triggers it, who can use it, and what it runs |
| [Webhook Listeners](/api-reference/configuration-api/tool-studio/webhook-listeners/overview)               | `/platform/extensions/listeners`                | Receives events from an external system            |
| [MCP Servers](/api-reference/configuration-api/tool-studio/mcp-servers/overview)                           | `/platform/extensions/mcp-workspace/servers`    | A remote MCP server and its tools                  |

Each collection supports list, create, get and update. There is no delete.

## Build in order

A resource has to exist before anything that refers to it. Create them in this order, and keep the `id` from each response:

```
HTTP connector → HTTP action or script → compound action → data type → conversational process → plugin
webhook listener → system plugin
HTTP connector → MCP server
```

Send writes one at a time. Writes aren't transactional: if a later step fails, the resources you already created stay, so reuse them or clean them up in Tool Studio.

## `id` or name?

Most references take the other resource's `id`. Two take a name:

| Field                                                                                | Send                                                  |
| ------------------------------------------------------------------------------------ | ----------------------------------------------------- |
| `connector_id` (HTTP actions, MCP servers)                                           | the HTTP connector's `id`                             |
| `action_id` and `dynamic_lookup` (`http_action`, `compound_action`, `script_action`) | the action's `id`                                     |
| `conversational_plugin_body_config_uuid`                                             | the conversational process's `id`                     |
| `listener_id`                                                                        | the webhook listener's `id`                           |
| `action_name` inside compound action `workflow_steps`                                | the action's `signature.name`                         |
| A custom slot or argument type, `{"custom": ...}`                                    | the data type's `name`, for example `u_PurchaseOrder` |

A `400` with `REFERENCE_VALIDATION_FAILED` usually means a name was sent where an `id` belongs, or the other way around.

## How updates work

`PATCH /platform/extensions/<collection>/{id}` takes a JSON merge-patch:

* A field you leave out keeps its stored value, and `null` clears it.
* Nested objects merge field by field.
* Lists of objects merge row by row instead of being replaced. To remove or reorder rows, read the document and send the whole list back.
* Lists of strings, Data Mapper expressions (`*_bender`, `input_mapper`, `output_mapper`), JSON schemas, `workflow_steps` and `resolver_methods` are replaced whole.
* Fields that are alternatives to each other, such as `allow_all` and `allow_emails`, switch when you send the other one. Sending both is rejected.
* Send `version.num` from your last read with a `change_description`. If the document changed since then, the update returns `409`; read it again and retry.

Data Mapper fields in these collections take the short form, for example `{"po_number": "data.po_number"}`. Rules such as `condition`, `skip_execution_condition` and `allowlist_rule` are DSL. See [MWDSL & Bender](/api-reference/configuration-api/mwdsl-and-bender).

## Reading

* Reads return every field. Unset fields come back as empty values: `""`, `0`, `false`, `[]` or `null`.
* Secrets come back redacted and can't be read back.
* To copy a resource, remove `id`, `updated_time`, `updated_by` and `version.num` from the read body before you send it as a create.

There is no endpoint that runs an action or a plugin. After you create one, test it in Tool Studio, or try the plugin in chat as someone on its allow list.