Skip to main content

MCP quickstart

This guide connects an AI client to FieldTask and creates a job from a plain-language request. You need to be an owner or admin of a FieldTask organization with API access enabled, and an AI client that supports remote MCP servers, such as Claude.

1. Connect your client​

  1. In FieldTask, open Settings → API connections. If you see Request API access, request it and wait for FieldTask to enable it.
  2. Under AI assistants, choose your client, for example Connect Claude, and follow the steps shown.
  3. When the FieldTask sign-in page opens, sign in, choose your organization, choose Read + create under Access you grant, and select Allow access.

See Connect an AI client for other clients and command-line setup.

2. Check the connection​

Ask your client:

What jobs are scheduled for today in FieldTask?

The client calls fieldtask_list_jobs and answers from your data. If it can't find FieldTask's tools, check that the connection appears under Active connections in Settings → API connections, and refresh the client's tools.

3. Create a job​

Ask, using a real customer from your organization:

Create an unscheduled job called "MCP quickstart test" for Example Property Services.

The client looks up the customer, may ask you to confirm which customer or location you mean, then calls fieldtask_create_job. Its wording varies, but you can check the effect:

  • The job appears under Jobs in FieldTask, for the customer you named.
  • The tool result includes the new job's data.id and an operation_id.

4. Confirm the outcome​

Ask:

Check the operation for that job.

The client calls fieldtask_get_operation. The result shows local_status: "committed", meaning the job is saved.

If a create ever times out or looks uncertain, ask the client to check like this before asking it to try again.

Next steps​