Model Context Protocol

Monitoring MCP server.
Your monitoring,
inside your agent.

Sentinelboard monitors your websites, APIs, cron jobs and servers. Connect Claude Code, Cursor, Codex or VS Code to it. Ask what is down, investigate incidents and server vulnerabilities, and set up monitors from a prompt.

Connect your client

Free tier available. No credit card required.

Hosted · nothing to installStreamable HTTPGitHub ↗

› Is anything broken right now?

⏺ sentinelboard_list_metrics

⏺ sentinelboard_list_incidents

⏺ sentinelboard_get_server

Two things need attention:

● checkout-api has returned 503 for 12 minutes. An incident is open.

● db-01 disk is at 91%, above your warning threshold.

Your other 23 monitors are healthy. Want me to acknowledge the incident?

Example agent session · Illustrative data

Why use the Sentinelboard MCP server

Your whole workspace, in context

Dashboards, monitors, history, incidents, error groups, servers, PostgreSQL and vulnerability findings, all as structured tools your agent can use.

Changes without surprises

Create and update monitors, dashboards and alerts with idempotent retries. Archiving is reversible and there are no permanent-delete tools.

Tokens you control

Dedicated read-only or read/write tokens with an expiry date. Revoke any token at any time without touching your other credentials.

Ask, don't click

Monitoring questions,
answered from live data.

Your agent picks the right tools, reads the current state and answers in plain language. Use it during an incident, while you deploy, or to set up monitoring for a new project without leaving your editor.

  • “Which of my monitors are failing right now, and since when?”

    Read-only token
  • “Summarize open incidents from the last day and resolve the ones I’ve already fixed.”

    Read & write token
  • “Which servers have vulnerabilities I should patch first?”

    Read-only token
  • “Add uptime and SSL monitoring for staging.example.com to my Production dashboard.”

    Read & write token
  • “Is example.com reachable from outside, and when does its certificate expire?”

    Read-only token
  • “Show the newest application errors with their stack frames.”

    Read-only token

Tool catalog

Read everything.
Change what you allow.

All tool names start with sentinelboard_. Agents discover the full schemas through MCP, so new capabilities appear without changing your configuration.

Read tools cover error tracking, server and PostgreSQL monitoring and the server security checker.

Read tools

Available with every token.

Workspace & dashboards

  • get_overview
  • list_dashboards
  • get_dashboard

Monitors & history

  • list_metrics
  • get_metric
  • get_metric_history

Alerts & incidents

  • list_alerts
  • list_incidents
  • get_incident

Application errors

  • list_error_logs
  • get_error_log

Servers & PostgreSQL

  • list_servers
  • get_server

Server vulnerabilities

  • list_security_findings
  • get_security_finding

Live public checks

  • check_web_availability
  • check_ssl_certificate

Write tools

Require a read & write token. Only resources you own can be changed.

Monitors

  • create_metric
  • update_metric
  • set_metric_enabled
  • archive_metric
  • restore_metric

Events & reminders

  • record_metric_event
  • complete_reminder
  • postpone_reminder

Dashboards

  • create_dashboard
  • update_dashboard
  • add_metric_to_dashboard
  • move_metric
  • remove_metric_from_dashboard

Alerts, incidents & errors

  • create_alert
  • update_alert
  • update_incident
  • dismiss_error_group

Monitors an agent can create

Connect in two minutes

One URL.
One token.

  1. 01

    Create a token

    Sign in and open Settings → AI agents. Start with Read only; choose Read and write when your agent should make changes.

  2. 02

    Store it securely

    Put the token in the SENTINELBOARD_MCP_TOKEN environment variable or your client's secret store. Never paste it into a chat.

  3. 03

    Add the server

    Use the configuration for your client, restart it, and ask your agent for a Sentinelboard overview to verify the connection.

Let your agent set it up

Point your agent at the Sentinelboard MCP skill. It explains the configuration for your client, keeps the token out of the conversation and verifies the connection with read-only calls.

Add this to your project’s .mcp.json. Claude Code expands ${SENTINELBOARD_MCP_TOKEN} from your environment, so the secret stays out of the file.

.mcp.json
{
  "mcpServers": {
    "sentinelboard": {
      "type": "http",
      "url": "https://api.sentinelboard.com/mcp",
      "headers": {
        "Authorization": "Bearer ${SENTINELBOARD_MCP_TOKEN}"
      }
    }
  }
}

Add the server to your global or project mcp.json. Cursor reads the token from the SENTINELBOARD_MCP_TOKEN environment variable.

~/.cursor/mcp.json
{
  "mcpServers": {
    "sentinelboard": {
      "url": "https://api.sentinelboard.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:SENTINELBOARD_MCP_TOKEN}"
      }
    }
  }
}

VS Code asks for the token in a masked prompt the first time the server starts and stores it securely.

.vscode/mcp.json
{
  "inputs": [
    {
      "type": "promptString",
      "id": "sentinelboard-token",
      "description": "Sentinelboard MCP token",
      "password": true
    }
  ],
  "servers": {
    "sentinelboard": {
      "type": "http",
      "url": "https://api.sentinelboard.com/mcp",
      "headers": {
        "Authorization": "Bearer ${input:sentinelboard-token}"
      }
    }
  }
}

Codex reads the token from the environment variable you name. The config file never contains the secret.

~/.codex/config.toml
[mcp_servers.sentinelboard]
url = "https://api.sentinelboard.com/mcp"
bearer_token_env_var = "SENTINELBOARD_MCP_TOKEN"

Use the Streamable HTTP transport and send your token in the Authorization header on every request. The server is stateless and needs no session affinity.

Any MCP client
Transport:  Streamable HTTP
URL:        https://api.sentinelboard.com/mcp
Header:     Authorization: Bearer sb_mcp_…

More clients, the registry manifest and troubleshooting are in the sentinelboard-mcp repository on GitHub.

Built for agent access

Useful to agents.
Safe for your account.

Scoped to you

Every lookup is limited to the token’s user. Shared dashboards are readable; only resources you own are writable.

Nothing is destroyed

No permanent-delete tools. Archived monitors can be restored, and dashboard removal keeps the monitor.

Every call audited

Tool arguments and results are logged for 30 days, with credential fields such as headers, tokens and passwords filtered out.

Secrets stay hidden

Tokens are hashed at rest. Monitor credentials and raw error payloads are never returned to the agent.

Live availability and SSL checks accept public targets only; private and internal addresses are blocked.

The practical details

MCP server FAQ

Clients, authentication, permissions and what agents can change.

What is an MCP server?

The Model Context Protocol (MCP) is an open standard that lets AI agents use tools from other applications. The Sentinelboard MCP server gives a compatible agent structured access to your dashboards, monitors, incidents, errors and servers, so it can answer monitoring questions with your live data instead of guessing.

Which AI clients can I connect?

Any MCP client that supports remote Streamable HTTP servers with a custom Authorization header. That includes Claude Code, Cursor, VS Code, Codex and many agent frameworks and SDKs. There is nothing to install: the server is hosted by Sentinelboard.

How does authentication work?

Create a dedicated MCP token in Settings → AI agents. Choose Read only or Read and write, and an expiry between 30 and 365 days. Each token is shown once, stored only as a hash, and can be revoked independently of your password, session and ingestion keys.

Can an agent delete my monitors or data?

No. There are no permanent-delete tools. An agent with a read/write token can archive a monitor, which is reversible, or remove it from a dashboard without deleting it. Read-only tokens cannot change anything.

Which monitors can an agent create?

Website availability, API response, SSL certificate, domain expiration, heartbeat, numeric, stock, crypto, RSS, reminder and periodic event monitors. Inbox, error ingestion, OpenRouter and server monitors can be read but are set up in Sentinelboard, because their setup involves credentials or an agent installation.

Will retries create duplicate monitors?

No. Creation and event-recording tools require an idempotency key. When an agent retries the same action with the same key, Sentinelboard returns the original result instead of creating a duplicate.

What does Sentinelboard record about agent activity?

Every tool call is audited with its arguments and result so you can review what an agent did. Credential-bearing fields such as headers, bodies, tokens and passwords are filtered from the audit log, which is retained for 30 days.

Is the MCP server open source?

The server runs as part of the hosted Sentinelboard service. Setup guides, client configuration examples, the registry manifest and the agent skill are published in a public GitHub repository.

Give your agent eyes on production.

Uptime, incidents, errors and servers in one workspace, available to you and to the AI tools you already use.

View all features

Free tier available. No credit card required.