> ## Documentation Index
> Fetch the complete documentation index at: https://docs.protodesk.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Build with your AI

> Give your AI the right documentation, connect your workspace, and build a working integration.

Start with the outcome you want: connect your AI to your workspace, or ask it to write an integration for your product.

## Choose your path

| Your goal | Use | First step |
| - | - | - |
| Ask your AI to find conversations, suggest replies, translate text, or propose articles | Protodesk MCP | Follow the [setup instructions for your AI](/mcp/connect) |
| Build customer sync, a support workflow, or a webhook receiver in your own application | REST API | Give your coding assistant the starter prompt below |

For MCP, the server URL is `https://mcp.protodesk.io/mcp`. Follow your client's setup instructions and sign in through OAuth when supported. Client support varies; the setup page describes API-key options where needed.

## Connect and check your workspace

After completing [MCP setup](/mcp/connect), paste this into your AI:

```text theme={null}
Use Protodesk workspace_context to check my connection.
Tell me the workspace name, workspace ID, and available permissions.
Then show up to five recent conversations if my permissions allow it.
Keep this check read-only. If a permission is missing, explain which one.
```

Check the workspace before continuing. For your next task, include the outcome and any limits, such as “Draft a reply to this conversation and show it to me before sending.” See [use case examples](/mcp/use-cases) for more prompts.

## Build an integration with a coding assistant

Replace the bracketed fields, then paste this prompt into your coding assistant. It works as a project brief for an assistant that can read documentation and edit your code.

```text theme={null}
Build a Protodesk integration in this project.

Goal: [describe the workflow and what success looks like]
Language and framework: [your stack, or inspect this repository]

Read these official sources before choosing endpoints or fields:
- Documentation index: https://docs.protodesk.io/llms.txt
- REST API specification: https://protodesk.io/openapi.yaml
- First request: https://docs.protodesk.io/quickstart
- Errors and retries: https://docs.protodesk.io/guides/errors-and-retries
- Testing: https://docs.protodesk.io/developer/testing

Use https://api.protodesk.io/v1 as the REST base URL.
Use only documented endpoints and identify the minimum required scopes.
Keep PROTODESK_API_KEY in a server-side environment variable.
Never put a real key in source code, client-side code, logs, or this chat.

Implement a read-only connection check with GET /me and verify the expected
workspace ID before enabling writes. Follow cursor pagination where needed.
Handle rate limits, timeouts, and supported idempotent retries without
creating duplicate actions. Do not assume every write supports idempotency.

Add focused tests using mocked responses, plus setup instructions and an
environment-variable example containing placeholders only. Do not send
messages, create live data, or consume AI credits during initial setup.

Explain what you implemented, how to test it, and any missing capability.
If the documentation does not support a requirement, tell me rather than
inventing an endpoint or an official SDK.
```

If your assistant cannot fetch a link, provide the relevant page using its **Copy page** option, or download and attach the API specification. The [documentation index](https://docs.protodesk.io/llms.txt) links to Markdown versions of the pages.

## Supply credentials locally

Create a scoped API key using [Make your first request](/quickstart). Add it to your application's secret manager or local environment outside the AI conversation. An MCP OAuth connection authorizes MCP access; it does not replace the API key used by your REST application.

## Verify the result

Run the read-only connection check first. Confirm that the workspace and scopes match your intention, then follow [Testing your integration](/developer/testing) before enabling writes. Review the [production checklist](/developer/production-checklist) before deploying.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.