AI assistants (MCP)

Your AI assistant, such as Claude Code, Cursor or VS Code, can answer questions about your Monitors and act on them. Ask it things like:

• “Is checkout down, and since when?”
• “Add a heartbeat for my nightly backup”
• “What was our uptime last month?”

It reaches You Up? through its MCP server, at https://api.uup.dev/mcp. MCP is the way AI assistants connect to services like You Up?. The assistant’s key is an ordinary API token, so it does what the token allows, in the Projects the token reaches, and follows the same rules as the dashboard.

Which kind of token to make

Start with Read-only. The assistant can then read your Monitors, their results, uptime, Incidents and alerts, and change nothing.

A Full token also lets it create, change, pause and resume Monitors, acknowledge and resolve Incidents, and snooze Checks. It never deletes anything, whatever the kind.

The fast way

The owner and admins can connect an assistant from the dashboard:

1. In the sidebar, open AI assistants.
2. Press Connect an assistant, and choose your assistant, what the token reaches, its kind and when it expires.
3. Press Connect, then Copy prompt, and paste the prompt into your assistant. It sets itself up from it.

The token is named after the assistant and kept on the API tokens screen, where you revoke it. It can be copied only while the dialog is open.

By hand

Make an API token and copy it. Then paste this prompt into your assistant, with your token in place of uup_api_…:

Connect to uup, my uptime monitoring service, through its MCP server. URL: https://api.uup.dev/mcp (Streamable HTTP). Header: Authorization: Bearer uup_api_…. Add it as a server named "uup" in my user-level MCP settings, never in a file committed to a repository. Then list uup's tools and call get_status to check that it works.

Or set your assistant up yourself. Put your token in place of uup_api_…, and keep the setup out of any file you commit to a repository.

Claude Code

Run this in a terminal.

claude mcp add --transport http --scope user uup https://api.uup.dev/mcp \
  --header "Authorization: Bearer uup_api_…"

Cursor

Add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "uup": {
      "url": "https://api.uup.dev/mcp",
      "headers": { "Authorization": "Bearer uup_api_…" }
    }
  }
}

VS Code

Run MCP: Open User Configuration, and add this to mcp.json:

{
  "servers": {
    "uup": {
      "type": "http",
      "url": "https://api.uup.dev/mcp",
      "headers": { "Authorization": "Bearer uup_api_…" }
    }
  }
}

Windsurf

Add this to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "uup": {
      "serverUrl": "https://api.uup.dev/mcp",
      "headers": { "Authorization": "Bearer uup_api_…" }
    }
  }
}

Zed

Add this to your Zed settings.json:

{
  "context_servers": {
    "uup": {
      "url": "https://api.uup.dev/mcp",
      "headers": { "Authorization": "Bearer uup_api_…" }
    }
  }
}

Codex

Add this to ~/.codex/config.toml:

[mcp_servers.uup]
url = "https://api.uup.dev/mcp"
http_headers = { "Authorization" = "Bearer uup_api_…" }

claude.ai

1. Open Customize, then Connectors, and choose Add custom connector. On Team and Enterprise, an owner does this in Organization settings, then Connectors.
2. Name it uup. For the server URL, paste https://api.uup.dev/mcp. Under Authentication, choose No sign-in.
3. Under Request headers, add Authorization with the value Bearer uup_api_….
4. Choose Add, then turn on uup in a chat's tools menu.

Request headers shows only for some Claude organizations. Without it, claude.ai cannot send the token; set up Claude Desktop instead.

On Team and Enterprise, everyone in your Claude organization shares this token, so choose Read-only.

Claude Desktop

Needs Node.js.

Add this to claude_desktop_config.json:

{
  "mcpServers": {
    "uup": {
      "command": "npx",
      "args": ["mcp-remote", "https://api.uup.dev/mcp", "--header", "Authorization:${AUTH_HEADER}"],
      "env": { "AUTH_HEADER": "Bearer uup_api_…" }
    }
  }
}

ChatGPT

ChatGPT cannot connect yet. It cannot use an API token, so there is no way to set it up today.

The tools

These are the tools your assistant can call. A Read-only token is offered the ones that read; a Full token is offered them all.

Tool	What it answers	Reads or writes
get_status	What is not ok now	Read
list_projects	List Projects	Read
list_monitors	Find Monitors	Read
get_monitor	Read a Monitor	Read
get_uptime	Read uptime	Read
list_incidents	List Incidents	Read
get_incident	Read an Incident	Read
list_runs	List runs	Read
get_response_times	Read response times	Read
list_alert_rules	List who is alerted	Read
list_deliveries	Did the alert go out?	Read
get_plan	Read the Plan	Read
list_locations	List locations	Read
create_monitor	Create a Monitor	Write
update_monitor	Change a Monitor	Write
pause_monitor	Pause a Monitor	Write
resume_monitor	Resume a Monitor	Write
acknowledge_incident	Acknowledge an Incident	Write
resolve_incident	Resolve an Incident	Write
snooze_check	Snooze a Check	Write
end_snooze	End a snooze	Write

When You Up? adds a tool, your assistant sees it the next time it connects. A tool is changed or removed only as API changes says.

Limits

Every tool call counts against the API’s limits, the same as a request your own code sends. Errors and limits says how many a minute you may make.