0%

Observe

Monitors tools for assistants (MCP)

The 5 tools an assistant can call for Monitors over MCP: what each one does, the scopes it needs, its arguments and an example call.

Scheduled checks on a connection: each monitor's runs, health, uptime and latency over time, and how long their data is kept. Create a monitor for a connection, change or delete it, list its runs, and read its summary and series for charts. Use it to know when something you depend on goes down and how often it has. An assistant calls these tools over MCP; the same operations over HTTP are in the Monitors API. Tools for assistants lists every product, and MCP shows how to connect an assistant.

connections_list_monitors

List monitors. An account's monitors, newest first, each with its state (up, down, unknown, paused), target kind, interval and where it checks from. Narrow by connection_id, agent_id, category or tags (key:value). org_id is the account (accounts_get_me names yours).

Scopes: mcp:read, connections:read · reads only

Argument	Type	What it is
`account_id`	string	The account's id; an alias of org_id. Leave both out for the caller's own account.
`agent_id`	string	Only the monitors that run on this agent, "agt_" and a ULID: those it declared and those of its connections.
`category`	string	Only this category's monitors; "none" for the uncategorised.
`connection_id`	string	Only this connection's monitors.
`org_id`	string
`page_size`	integer	Page size: 0 is 50, above 200 is 200.
`page_token`	string	The previous answer's next_page_token; opaque. Reused with other arguments it is INVALID_ARGUMENT.
`tag`	list of string	Only the monitors carrying every one of these tags, each "key:value"; at most 10.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "connections_list_monitors",
    "arguments": {
      "page_size": 20
    }
  }
}

connections_get_monitor

Get a monitor. One monitor: what it checks, from where, how often, its failure threshold and its current state.

Scopes: mcp:read, connections:read · reads only

Argument	Type	What it is
`account_id`	string	The account's id; an alias of org_id. Leave both out for the caller's own account.
`monitor_id`	string
`org_id`	string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "connections_get_monitor",
    "arguments": {
      "monitor_id": "mon_01JD8Z3QK4W5X6Y7Z8A9B0C1D2"
    }
  }
}

connections_list_monitor_runs

List a monitor's checks. A monitor's check runs, newest first: when each ran, from where (cloud or an agent), pass or fail, latency and the kind of error. Never a response body. This is the evidence for whether something was really down.

Scopes: mcp:read, connections:read · reads only

Argument	Type	What it is
`account_id`	string	The account's id; an alias of org_id. Leave both out for the caller's own account.
`from`	string	RFC 3339: only runs at or after this time.
`monitor_id`	string
`org_id`	string
`page_size`	integer	Page size: 0 is 50, above 200 is 200.
`page_token`	string	The previous answer's next_page_token; opaque. Reused with other arguments it is INVALID_ARGUMENT.
`status`	string	Only runs with this status: "failed" (failed against the target), "passed", or empty for every run (the list contract's `status` filter).

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "connections_list_monitor_runs",
    "arguments": {
      "monitor_id": "mon_01JD8Z3QK4W5X6Y7Z8A9B0C1D2"
    }
  }
}

connections_get_monitor_summary

A monitor's health. A monitor's health now: uptime over 24 hours, 7 and 30 days, latency percentiles and its last failure.

Scopes: mcp:read, connections:read · reads only

Argument	Type	What it is
`account_id`	string	The account's id; an alias of org_id. Leave both out for the caller's own account.
`monitor_id`	string
`org_id`	string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "connections_get_monitor_summary",
    "arguments": {
      "monitor_id": "mon_01JD8Z3QK4W5X6Y7Z8A9B0C1D2"
    }
  }
}

connections_get_monitor_series

A monitor over a time range. A monitor's time range in equal buckets: runs, passes, failures, uptime and latency percentiles per bucket, and the failures by kind. At most 500 buckets, within the days runs are kept.

Scopes: mcp:read, connections:read · reads only

Argument	Type	What it is
`account_id`	string	The account's id; an alias of org_id. Leave both out for the caller's own account.
`bucket_secs`	integer	Seconds per bucket, at least 60; 0 picks a width that gives about 100 buckets. More than 500 buckets is INVALID_ARGUMENT.
`from`	string	RFC 3339; `to` is now and `from` 24 hours before it when empty. The range may reach back as far as runs are kept.
`monitor_id`	string
`org_id`	string
`to`	string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "connections_get_monitor_series",
    "arguments": {
      "monitor_id": "mon_01JD8Z3QK4W5X6Y7Z8A9B0C1D2"
    }
  }
}

Sensors

Previous Page

Incidents

Next Page