Model Context Protocol (MCP)

The Model Context Protocol (MCP) is a standard way for AI apps to connect to external systems. Catalyst uses MCP to let a supported AI app work with TrialStack on your behalf.

MCP provides the connection, while TrialStack remains in control. Your TrialStack access determines which organizations, records, and actions are available.

Most people do not need to choose tools themselves. The AI app chooses a tool from your request, and Catalyst checks whether it is allowed.

For browser-native tools that reflect the current signed-in Studio page, see WebMCP. WebMCP does not replace this remote Catalyst connection and cannot persist form changes.

Choose an organization

When you sign in, you may select a default organization. You are not locked to it. The same connection can use any organization that appears in your TrialStack access list.

If your request does not name an organization, ask:

Which TrialStack organizations can I access?

Then tell the app which organization to use. Catalyst checks your membership and permissions again for each request.

Understand access results

What you see	What it means	What to do
Sign-in required	Your connection is missing or has expired.	Reconnect Catalyst and sign in again.
Access denied	Your role cannot use that organization/action.	Ask a TrialStack administrator for access.
No results or total: 0	The request worked but found no matching data.	Check the organization and search details.
Operation is not available	Catalyst does not expose that action.	Use a supported TrialStack workflow instead.

An empty result is not an authentication failure. Check the selected organization before reconnecting.

Review and approve changes

Read-only questions can run immediately. A request that changes TrialStack always follows a separate preview and approval process:

1. Ask Catalyst to prepare the change.
2. Review the organization, record, proposed details, and expiry.
3. Correct the request if anything is wrong. Catalyst will create a new preview.
4. Approve the correct preview in a new message.
5. Catalyst applies the approved change and returns the affected record or run.

Preparing a change does not apply it. Do not approve a preview until you have checked its full scope.

For security, a preview expires after ten minutes and can be used only once. If it expires or your request changes, ask Catalyst to prepare it again.

Technical tool reference

The names below help administrators and developers understand logs or diagnose a conversation. Most users can describe the task in ordinary language.

Task	Catalyst tools
List available organizations	list_organizations
Find or open records	search, fetch
Find a supported API operation	search_operations, describe_operation
Run a read-only operation	call_read_operation
Create a change preview	prepare_change
Apply an approved change	execute_create_change, execute_update_change, and other effect-specific tools

Catalyst does not expose account administration, billing, credentials, webhooks, hard deletion, or data-purge operations. An unavailable operation is an intentional boundary.

If something goes wrong

I cannot connect

Confirm that your app uses this address:

https://api.trialstack.com/mcp

Then reconnect and complete the TrialStack sign-in flow. Opening the address as a normal web page is not a complete connection test.

I cannot see an organization

Ask a TrialStack administrator to check that Catalyst is enabled for the organization and that your current role includes Catalyst access.

I cannot find an expected trial

Ask Catalyst to list your organizations and confirm which one it is using. A successful empty result usually means the selected organization has no matching records.

My browser shows a localhost error after sign-in

Some command-line apps briefly use a local browser callback. If the app says authentication succeeded, verify the connection with /mcp or a read-only request before signing in again.

My approved change returned an error

Ask Catalyst to read the affected record before trying again. The change may have completed even if the final response failed. Checking first prevents an accidental duplicate.

My preview cannot be approved

The preview may have expired, already been used, or no longer match your request. Ask Catalyst to prepare a new preview and review it again.

See Agent Skills to learn how supported AI apps receive Catalyst instructions.