MCP tool execution

The check monitors that MCP server your agent takes tools from responds and the needed tool is present. Server is usually alive: more often its contents change — tool was renamed or removed from output. A regular HTTP check will see working address and notice nothing.

What exactly to check

You choose one level, not a set of checkmarks: they are nested — "tool exists" without "server responds" doesn't happen. First two levels execute nothing on the server and cost the same, so start with the second: it's what the check is for.

Level What is checked Cost
Server responds Server accepted request and response is parsed. Catches unavailable server, wrong address, and rejected token, and at explicitly selected protocol version — server moving to another. Free
Tool is on server Specified tool is in server's tools list. Catches renamed, removed from output, or rights-hidden tool — something that happens noticeably more often than server itself going down. Free
Tool executes Tool is called with your arguments and responds without error. Catches what list doesn't see: tool is there but what it calls is broken — database, external API, rights. Real call on each check
Third level actually executes the tool. Tool execution is an action. Monitoring that sends email every five minutes or creates ticket will cause quiet disaster, and will look "working" itself. So level is off by default, and choose only read tool for it — search, read, information. We can't tell read tool from write tool by description, so choice is yours.
Protocol version — automatically or yours. By default we figure out version ourselves on each check: any server supports will do. This is enough if you just need the tool to respond. But auto-detection silently follows the server: it moved to new revision, check adapted and shows "works" — but your application built for old version is already broken. If your code is tied to specific version, select it: then moving to another becomes failure you learn about. Not supporting selected version, server will name its versions, and we'll show them in check reason.
We monitor HTTP only. MCP servers can be local — those the editor or agent runs on your machine and talks to through stdin/stdout. No one outside can reach such server, including us. You can monitor server accessible by address from internet.
Outdated transport not supported. First protocol revision, 2024-11-05, worked differently: server held separate event stream and separate address for messages. It was deprecated back in spring 2025, and it's not in versions list. If your address uses it, check will say so directly — address doesn't accept normal MCP requests.

Form fields

Field What it means What to specify Example
Name Name of the check. Specify which server and what for is important to you. MCP for database search
Group Allows assigning task to project or section. Select a group if you want to keep project tasks together. Main project
Server address MCP endpoint — the one you put in agent settings. Enter full address with path. Usually path ends in /mcp, but it can be anything. https://example.com/mcp
Protocol version MCP revision your code is built for, or "Automatically" — any server supports. Leave "Automatically" if only the tool responding matters. Select same version as your client if you want to learn about server move to another. Automatically
Token Bearer token if server is closed with auth. Leave empty if server is open. When updating task empty field means "keep previous token"; when moving to different host previous token doesn't carry — we won't send it to new host. Token from server settings
What to check Check level from described above. Take the second: it executes nothing on server, like the first, and answers the question the check is for. Third — only for read tool. Tool is on server
Tool Name of tool we're monitoring. Asked from second level on. Specify exactly as server provides in tools list: check is exact. Same name the tool is called with, not its description. search_documents
Arguments What to call tool with. Asked only at third level. JSON object as tool expects arguments. Pick harmless query the tool will definitely answer. Empty field — call without arguments. {"query": "test"}
Timeout How much time allowed for entire check, at any level. Check longer than specified is failure: server that responds but slow is also broken. Base on how server usually responds, with margin. 30 seconds
Interval How often to run the check. On first two levels — as convenient, reasonable frequency — five minutes. On third — from 15 minutes: each check calls the tool. 5 minutes
Notification channels Where to send messages about issue and recovery. Check channels where you actually read notifications. Telegram and email
Reports Periodic summaries for the task. Enable if you want overview without manual panel access. Weekly report

When alert comes

Alert comes if selected check fails:

  • server unavailable, responded with error, or rejected token;
  • server doesn't support selected protocol version — check reason lists what it supports;
  • address responds not MCP server or server on outdated transport;
  • specified tool not in list — on second and third levels;
  • tool executed with error or requested input instead of result — on third level;
  • entire check took longer than timeout — at any level.

When check is skipped

If server asked to check less often, task doesn't fail or recover — check is skipped, status stays the same. This is the only such case: frequency limit means our own pushiness, not your server's problem. Everything else — silent network, server error, rejected token — for this task means you can't use the tool right now.

One task — one tool

Monitoring three tools of one server — that's three tasks. So each has its uptime and history: mixing questions in one task, you get uptime you don't understand whose it is. If tools are equally important, group tasks.

Notification settings

For each notification channel in a task additional parameters can be set. They apply only to DOWN events.

Parameter What it does
Enabled Allows or disables sending notifications via this channel for this task.
Delay for PRO Delays first DOWN notification for selected time (up to 12 hours). Convenient to avoid alerts for short-term failures.
Repeat every PRO Periodically repeats notification while task stays in DOWN status (up to 1 hour between repeats). Suitable when important not to miss prolonged failure.
"Delay" and "Repeat" parameters don't affect UP events and recovery notifications.
To the list of task types