Give your AI the ratel's keys
Die Ratel now speaks the Model Context Protocol. Connect Claude, ChatGPT, Cursor or your own agent and say in plain language what should happen: block TikTok after bedtime, show what was blocked this week, or build a dashboard for the school network.
- Works with any MCP client
- Your key, your account, your rules
- Stats ready for charts
Everything you do in the dashboard, in one sentence
The agent works with the same rules, roles and plan limits as the dashboard. It can only do what your key may do.
Change rules
Categories, apps, YouTube, SafeSearch, your own allow and block lists, bedtimes, and the block page.
Build dashboards
Time series per hour and per day, blocks by category, top domains and a per-profile comparison, with a fixed schema.
Answer questions
"Why is YouTube blocked?" "Which device isn't talking to us?" The agent reads the policy and the activity log.
Chat with the Ratel
The chat tool talks to Die Ratel's own assistant, just like in the dashboard.
Ask for a chart, get a chart
get_stats and get_stats_by_profile return structured data with an output schema: an AI knows exactly which fields exist and builds the chart without guessing.
Stats are only as complete as each profile's retention setting keeps.
How to connect
The server lives at
https://mcp.dieratel.com/mcpCreate an API key
Sign in at my.dieratel.com, go to AI agents under Protection and create a key. API access is part of the Pro 100 plan.
Add the server
Paste the address and your key into your AI client (see below).
Ask something
Start with "Show me my profiles". The agent discovers the tools itself.
claude.ai
Customize > Connectors > Add custom connector. Paste the address and click Connect; a Die Ratel page asks for your key once:
https://mcp.dieratel.com/mcp Claude Code
In a terminal:
claude mcp add --transport http dieratel https://mcp.dieratel.com/mcp \
--header "Authorization: Bearer dr_live_..." Cursor, Windsurf and others
In the client's MCP settings (mcp.json):
{
"mcpServers": {
"dieratel": {
"url": "https://mcp.dieratel.com/mcp",
"headers": { "Authorization": "Bearer dr_live_..." }
}
}
} Claude Desktop
In claude_desktop_config.json, through mcp-remote:
{
"mcpServers": {
"dieratel": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.dieratel.com/mcp",
"--header", "Authorization:Bearer dr_live_..."]
}
}
} Anything else (curl)
The protocol is plain JSON over HTTP:
curl https://mcp.dieratel.com/mcp \
-H "Authorization: Bearer dr_live_..." \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' ChatGPT, OpenAI agents and other clients that support "Streamable HTTP" and an Authorization header work the same way.
The tools
35 tools, in groups:
Read
get_accountlist_profilesget_policylist_devicesget_activitysummarise_profilecheck_devicescheck_networkDashboards
get_statsget_stats_by_profilelist_categoriesChange rules
set_categoriesset_platformsset_youtubeset_safesearchblock_domainallow_domainremove_domainsset_blocked_tldsset_block_allset_block_pageapply_age_presetBedtime and profiles
set_bedtimeclear_bedtimecreate_profilerename_profileset_retentionDevices and setup
add_deviceremove_devicedevice_instructionsrouter_instructionsdraft_isp_emailTalk
chatrecommend_planbilling_statusSafe by design
One key, one account
A key only works on the account that created it. Revoke it in the dashboard and the agent is out immediately.
The same rules as you
Every request goes through the same API as the dashboard: roles, plan limits and validation apply unchanged.
No payments
An agent cannot pay, cancel or move your home network. That stays with a person in the dashboard.
Nothing stored
The MCP server keeps no data and no log of requests; it only passes them on.
Questions
What is MCP?
The Model Context Protocol is an open standard that lets AI applications use outside tools. Claude, ChatGPT, Cursor and many others support it.
Which plan do I need?
API access, and so the MCP server, is part of the Pro 100 plan.
Can the agent loosen my children's rules?
Yes, if you ask it to and the key has that role. Give a key to an agent the way you would give a password, and revoke it when you're done.
Does the AI see my family's browsing?
The agent can only read the activity log your plan keeps (which websites were looked up, not what was on the pages), and only for as many days as your plan keeps it. If you'd rather not share it, ask only for summaries, or set the profile's retention to "reports only" or "nothing".
Let the AI do the work
The ratel still bites. Now you can just tell it where, in plain language.