Connecting AI Agents to TrekMail Drive

Open TrekMail in Claude, choose Connect, sign in and approve access. You do not need to create an API token. To review permissions or disconnect, open AI Agents & API → Tokens → Connected apps.

Your own White Label brand: use the MCP address shown in your branded dashboard and name the connector after your brand. Add it as a custom connector; the public TrekMail listing signs in on TrekMail and does not replace your branded connection.

An AI agent can manage Drive files safely when you give it a narrow token and clear rules. This guide focuses on Drive-specific setup for MCP-compatible agents. For general MCP installation, see Connecting AI Agents (MCP).

Choose the agent's job

Start by naming the workflow, not the tool. Good examples:

• client-delivery-links creates public links for approved files.
• monthly-report-uploader uploads generated PDFs into Account Drive.
• mailbox-file-assistant searches one mailbox Drive.
• storage-monitor reads usage and add-on status.

A focused name makes token review and audit logs much easier later.

Create a scoped token

Open AI Agents & API → Tokens → Create token. For a Drive-only agent, select only Drive scopes. Most agents do not need purge.

Examples:

Copy the token once. TrekMail will not show it again.

Configure MCP

Set the ops token in your MCP environment:

TREKMAIL_BASE_URL=https://trekmail.net
TREKMAIL_API_TOKEN=tm_live_your_token
TREKMAIL_ALLOW_DESTRUCTIVE=false

This is a self-hosted MCP safety control, not a Drive feature switch. Keep TREKMAIL_ALLOW_DESTRUCTIVE=false for normal Drive work and turn it on only for a reviewed cleanup run that needs purge tools.

If the same agent also reads or sends email, add a message token separately:

TREKMAIL_MESSAGE_TOKEN=tm_msg_your_message_token

Drive tools do not use message tokens. Message tools do not use Drive scopes.

Test with harmless tools first

Ask the agent to call:

1. drive_spaces_list
2. drive_storage_summary
3. drive_browse_folder for space="account"

These calls confirm the token works without changing data. Then test a small upload in a temporary folder.

Give the agent safe operating rules

A practical system prompt for a Drive agent should include:

• Never purge files unless the user explicitly asks.
• Prefer Trash over permanent delete.
• Add expiry and download limits to public links.
• Return created share URLs immediately.
• Use Account Drive for team files and mailbox Drive only when the task names a mailbox.
• Check storage usage before uploading large files.

Audit after setup

Open AI Agents & API → Audit Log and filter by token name. Verify that the agent only called the expected Drive tools. If a token has too many scopes, revoke it and create a narrower replacement.