Skip to content

monitoring--observability.md

Latest commit

 

History

History
142 lines (138 loc) · 43.9 KB

monitoring--observability.md

File metadata and controls

142 lines (138 loc) · 43.9 KB
Raw

📡 Monitoring & Observability

Servers connecting to monitoring systems, logging platforms, or providing system/application performance metrics.

  • sami-abdul/t3-watch: Monitoring that sets itself up from the projects on your machine and wakes your coding agent only when something changes. It reads the projects and logins you tick (GitHub Actions CI, sites and APIs, cloud accounts, API-key vendors, scheduled jobs that should ping in), writes the checks, and runs them on its own clock as plain code, so checks cost no model tokens. A failing, recovering or unrunnable check wakes Claude Code, Codex or T3 Code through a signed webhook, and the agent can read the check history, leave notes and draft new checks. Keys are pasted on a local page, never in chat, and vendor calls only read. Install: npx @lfg-labs/t3-watch demo runs it on sample products without touching any account. Transport: stdio. Registry id ai.lfglabs/t3-watch. TypeScript, MIT.

  • ConnectorReport: Check whether an MCP connector actually works before your agent uses it. Returns dated, hands-on test results for 561 remote MCP servers: pass rate, sign-in method, gotchas and fixes, plus daily handshake status. Tools: list_connectors, get_connector_report, check_connector_status, compare_connectors, search_reports. Remote streamable-http at https://connectorreport.com/api/mcp, no auth, read-only. Registry id com.connectorreport/connectorreport.

  • MCP Canary: Inspect any remote MCP endpoint without calling its tools — transport (Streamable HTTP or legacy SSE), protocol version, server info, tool list with input/output schemas, and a side-effect class per tool from its MCP hints; plus plans and prices of the MCP Canary monitoring service. Tools: inspect_endpoint, get_pricing, both read-only. Remote streamable-http at https://api.mcpcanary.com/mcp, no auth. Registry id com.mcpcanary/inspect. Scheduled tool-call monitoring with assertions and alerts is set up with the MIT canary CLI: https://github.com/xakpc-dev-labs/canary.

  • Shopify Speed Check: Hosted remote MCP (https://tools.skaidigital.com/mcp, no auth) to check any Shopify store's PageSpeed and Core Web Vitals, find slow apps and scripts, and compare Shopify plan fees.

  • Blackman99/cronlag: Slot-by-slot lag report for GitHub Actions on: schedule workflows — which cron slot each run served, how late it was created, which slots never ran, cause checks on your side, and catch-up via workflow_dispatch with a guard that skips GitHub's late duplicate. Read-only MCP (cronlag mcp: schedule_report, catchup_plan, cron_explain) plus CLI and Action. Install: npx github:Blackman99/cronlag. Transport: stdio. Auth: optional GITHUB_TOKEN for private repos. License: MIT.

  • FluxProof: Monitors public pricing, policy and changelog pages and returns cited before-and-after evidence with source URLs, content hashes and change history. Free: 1 page checked daily for 30 days, no card. Remote: https://fluxproof.steellantern.com/mcp.

  • CronCool: Inspect scheduled jobs, execution history, durable workflow traces and webhook health, and trigger one configured job after confirmation. Hosted Streamable HTTP at https://mcp.cron.cool/mcp; OAuth account authorization required. Docs. Public connector configuration and workflow skills: Apache-2.0.

  • Flare: Official Flare server by Spatie for error tracking, performance monitoring and logs in Laravel and PHP apps. List and inspect errors with stack traces; resolve, snooze, assign and annotate them; search logs; find the slowest routes, queries, jobs, commands and external HTTP calls and their traces; create projects; and search the Flare docs. Remote streamable-http at https://flareapp.io/mcp, OAuth 2.1 with dynamic client registration (personal access token as a fallback). Docs: https://flareapp.io/docs/flare/general/our-mcp-server.

  • Oh Dear: Official Oh Dear server by Spatie for website monitoring. Check uptime, SSL certificates, broken links, mixed content, DNS, cron jobs and status pages for the sites in your Oh Dear team. Remote streamable-http at https://ohdear.app/mcp, OAuth 2.1 with dynamic client registration, no API key needed. Registry id app.ohdear/ohdear.

  • EventSend: Official EventSend server. Ask your AI about your product's events, delivery health and usage: search events, follow one customer's timeline, see why a Slack, Discord, Telegram or webhook notification failed, and check daily stats and quota. 9 read-only tools. Remote streamable-http at https://eventsend.io/mcp, OAuth 2.1 with dynamic client registration (read-only mcp_ API key for local clients). Free plan available. Registry id io.eventsend/eventsend.

  • stephen1137/phpray: Ask why a PHP site is slow, from recorded production traffic rather than source code. Always-on request tracing: every request a site serves is stored with wall time, CPU, memory, SQL queries as fingerprints with the literals stripped, outbound HTTP calls, PHP errors and N+1 patterns; a sample adds per-plugin time attribution. 10 read-only tools: site list, overview, slowest pages, slowest queries, component breakdown, errors, trace search, one trace end to end, before/after compare, and live alerts. The hosted endpoint answers with no key from a read-only demo account holding real WooCommerce traffic, so the tools can be tried before installing anything. Runs where PHP usually runs — the extension loads from your own php.ini and the WordPress plugin needs neither an extension nor a daemon, so shared hosting without root works. Install: claude mcp add --transport http phpray https://phpray.dev/mcp. Transport: http (bearer console token, optional) or stdio. Registry id dev.phpray/phpray. Go/C, Apache-2.0.

  • magna-nz/tallybook: Local-first cost ledger for coding agents — prices every Claude Code and Codex session already on disk and lets the agent ask what it is spending mid-session. 8 tools, 7 read-only: spend report with findings, per-finding advice with an applyable patch, spend by sub-agent type, session list and per-session turn breakdown, model-swap verdicts, the price table, and a rescan. Reads local transcripts only — no proxy, no API key, nothing leaves the machine. Transport: stdio. Install: brew install --cask magna-nz/tap/tallybook, or go install github.com/magna-nz/tallybook/cmd/tallybook@latest. Glama: https://glama.ai/mcp/servers/magna-nz/tallybook. Go, MIT.

  • openITCOCKPIT/openITCOCKPIT-MCP-Server: Official server for the openITCOCKPIT (Naemon-based) monitoring platform — host and service state, check and state history, log entries, downtimes, acknowledgements, monitoring engine stats, software inventory and pending security updates. 39 tools, 24 read-only; the 15 configuration-write tools are not registered unless OITC_ENABLE_WRITE_TOOLS=true. Tools take hostnames and template names rather than database IDs, and writes are scope-checked against the target container. Install: docker run -d -p 8000:8000 --env-file .env openitcockpit/mcp-server:0.1.0, or pip install .. Transport: http (bearer token) or stdio. Requires openITCOCKPIT 5.6+. Python/FastMCP, MIT.

  • GT-dinuo/server-ops-mcp: Server ops via MCP — log troubleshooting, CPU/memory/disk monitoring, code edit, Nginx & certificate management. Local or SSH-remote with two-step confirmation, command whitelist, and secret redaction. Install: npx -y server-ops-mcp. Registry id io.github.GT-dinuo/server-ops-mcp.

  • stcmain/whats-running-mcp: Ground truth for agent fleets — live coding-agent sessions (Claude Code, Codex, Aider, Cursor…), listening TCP ports, loaded launchd/systemd daemons, and system load, read straight from ps/lsof/launchctl/systemctl rather than stale docs or memory. Five read-only tools; read-only by construction — fixed binaries via execFile, no shell. macOS and Linux. Install: npx -y whats-running-mcp. Registry id io.github.stcmain/whats-running-mcp. MIT.

  • uptimemonitoring/examples: Agent-native uptime monitoring via MCP — create, inspect, and assert monitor health, list incidents, and manage monitors. Remote streamable-http at https://api.uptimemonitoring.com/mcp, OAuth 2.1 or API key, no install. Free 50-monitor tier. Registry id com.uptimemonitoring/mcp. Docs: https://uptimemonitoring.com/docs/mcp/.

  • triadgit/humanhours-mcp: Track agent ROI and enrich companies via MCP: hours and money saved, ROI reports, and per-hour wage data. Remote streamable-http at https://mcp.humanhours.dev/mcp, OAuth 2.1, no install, no API keys. Registry id dev.humanhours/humanhours. Docs: https://humanhours.dev/docs/mcp.

  • Tethral-Inc/AgentRegistry: Agent Composition Records MCP server for logging AI-agent interactions and querying behavioral profiles through friction, coverage, stable-corridor, failure-registry, and trend lenses. 21 tools. Install: npx -y @tethral/acr-mcp.

  • kame6493-del/mcp-diagnostics: Website diagnostics MCP server for DNS, SSL, HTTP headers, security audit grading, WHOIS, and technology-stack checks. Install: npx mcp-diagnostics. No API keys required.

  • us-all/datadog-mcp-server - Datadog observability MCP server with 165 tools across metrics, monitors, logs, APM, RUM, incidents, status pages, and fleet automation. 4 workflow Prompts and incident-triage-snapshot aggregation tool.

  • vdalhambra/siteaudit-mcp: Instant website audits MCP with 11 tools — full SEO (20+ checks), security headers + SSL, Lighthouse performance via Google PageSpeed, multi-site competitive comparison, broken link crawler, WCAG accessibility audit, Schema.org structured data validation, competitor gap analysis, robots.txt parser. No API keys. FastMCP 3.2, MIT.

  • aliyun/alibabacloud-observability-mcp-server: Facilitates seamless integration with Alibaba Cloud's observability products, offering tools for efficient log and application monitoring.

  • shibley/apistatuscheck-mcp-server: Live and historical availability for 285 public APIs and developer platforms across 30 categories, for telling a third-party outage apart from a bug in the caller's own code. get_api_status probes the endpoint at call time and returns that alongside the hourly monitor's value with its timestamp, so a single-region blip is visible as a disagreement rather than reported as an outage. Streamable HTTP, stateless, read-only, no auth: https://apistatuscheck.com/api/mcp. Tools: list_apis, search_apis, get_api_status, list_down_apis, get_uptime_history, list_categories. Registry id io.github.shibley/apistatuscheck. Note: GET returns 405 (stateless serverless, no held-open SSE stream); POST-based clients work.

  • outagedeck/mcp: Official hosted server for checking whether a cloud or SaaS vendor is down before debugging your own code. Live status, stack-wide verdicts, incident timelines, and independent 7–90 day uptime history for 172 vendors, read from their official feeds. Remote streamable-http at https://outagedeck.com/api/mcp; public tools are keyless and read-only. Registry id com.outagedeck/outagedeck-status. Docs: https://outagedeck.com/developers/mcp.

  • devhelmhq/mcp-server: Provides AI agents with tools to monitor uptime, manage incidents, configure alerting, and query third-party dependency status. Supports HTTP, DNS, TCP, ICMP, and heartbeat monitors.

  • enomoto11/aws-cost-notifier-mcp-server: Monitors daily AWS costs and generates GitHub Issue reports for cost analysis and comparison.

  • binalyze/air-mcp: Facilitates natural language interaction with Binalyze AIR's digital forensics and incident response capabilities through a Node.js MCP server.

  • hiyorineko/mcp-rollbar-server: Facilitates LLM interaction with Rollbar's error tracking data, offering comprehensive access to error details, deployments, and user information.

  • saarw/akhq-mcp-server: Facilitates AI assistants' integration with the AKHQ Kafka monitoring tool through an experimental MCP server.

  • Chazzychouse/weather-mcp-server: Provides weather forecasts and alerts using the MCP protocol with data from the National Weather Service.

  • Bigsy/shadow-cljs-mcp: Monitors shadow-cljs builds, providing real-time status updates and integration with LLMs for verifying build success after ClojureScript file edits.

  • ZephyrDeng/pprof-analyzer-mcp: A Go-based server for analyzing Go pprof performance profiles, offering tools for generating flame graphs and interactive profiling sessions.

  • srcgrp/sentry-mcp-server: Integrate Sentry error tracking into your development workflow by accessing release health data and issue details directly within your MCP-enabled environment.

  • yincongcyincong/VictoriaMetrics-mcp-server: Facilitates interaction with VictoriaMetrics through an MCP server interface.

  • Infisical/infisical-mcp-server: Integrate with Infisical APIs using the Model Context Protocol for managing secrets and projects.

  • scottlepp/tempo-mcp-server: Enables AI assistants to query and analyze distributed tracing data from Grafana Tempo using the Model Context Protocol.

  • cline/cline-community: Facilitates seamless issue reporting from Cline to GitHub by automating metadata extraction and providing a preview feature.

  • scottlepp/loki-mcp: Facilitates querying Grafana Loki log data using the Model Context Protocol (MCP) with a Go-based server.

  • groundlight/groundlight-mcp-server: Facilitates image analysis and query handling through detectors that leverage natural-language processing for Groundlight.

  • mrwadams/otx-mcp: Facilitates real-time interaction with AlienVault OTX for threat intelligence analysis and management through a comprehensive set of MCP tools.

  • bobtista/honeybadger-mcp: Facilitates AI-driven error analysis by bridging AI agents with the Honeybadger error monitoring service.

  • maito1201/cloudrun-logs-mcp: Facilitates AI assistants in accessing Google Cloud Run logs and service information through an MCP interface.

  • Lee-1024/mcp-glances: Integrates MCP with LLMs to analyze server monitoring data via Glances API, providing real-time system status feedback.

  • jalexspringer/impact-mcp-server: Deploy a remote MCP server on Cloudflare Workers with OAuth login and connect it to Claude Desktop for seamless tool integration.

  • JordanDalton/PodscanMcpServer: Podscan MCP Server enables seamless integration with Podscan's API for efficient podcast data retrieval.

  • querypie/querypie-mcp-server: QueryPie MCP server empowers administrators with real-time data visualization, resource monitoring, and security auditing capabilities.

  • xprilion/mcp-telemetry: Enhance your chat system's observability by tracing and analyzing conversations with LLMs using Weights & Biases Weave.

  • ghrud92/simple-loki-mcp: Facilitates AI-driven log analysis by interfacing with Grafana Loki logs via the Model Context Protocol.

  • ggilligan12/kibana-mcp: Facilitates AI-driven interaction with Kibana Security alerts by tagging, adjusting status, and retrieving alerts.

  • yuki9541134/mcp-redash: Facilitates SQL query execution and data source management through Redash API integration.

  • PovedaAqui/suzieq-mcp: Facilitates interaction with SuzieQ network observability via MCP tools for querying and summarizing network state data.

  • naveen09/mcp_pagerduty: Facilitates PagerDuty integration for on-call management queries via an MCP server, compatible with Claude.

  • brunoborges/jvm-diagnostics-mcp: Enhances JVM diagnostic tools with a user-friendly prompt and exposes them via an MCP server for local use.

  • gkhays/mcp-sbom-server: Generates a Software Bill of Materials (SBOM) using Trivy scans, outputting in CycloneDX format.

  • GeLi2001/datadog-mcp-server: Facilitates seamless interaction with Datadog's API for monitoring, dashboards, metrics, events, logs, and incidents.

  • co0ontty/rivers-radar-mcp: Leverage large language models to query, analyze, and manage website security monitoring data via MCP Tools.

  • opensensor/bn_cline_mcp: Facilitates binary analysis with Cline using Binary Ninja through an MCP server.

  • weekitmo/mcp-sentry: Node.js and TypeScript-based server integrating with Sentry for error tracking and monitoring via MCP and SSE protocols.

  • mottibec/otelcol-mcp: Facilitates dynamic configuration of OpenTelemetry Collectors through MCP tools, enabling efficient management of receivers, processors, and exporters.

  • zereight/sentry-mcp: Facilitates interaction with the Sentry API to retrieve issue and event details using the Model Context Protocol.

  • kitakou0313/grafana-tempo-mcp: A TypeScript-based MCP server implementing a simple notes system with tools for note creation and summarization.

  • prashantjha/mcp-datadog: Integrates Datadog with MCP for enhanced monitoring and observability.

  • samanhappy/mcphub: MCPHub consolidates multiple MCP servers into a single SSE endpoint, offering centralized management and real-time monitoring through an intuitive dashboard.

  • getsentry/sentry-mcp: Middleware MCP server enabling seamless interaction with Sentry through LLMs, supporting both remote and stdio transport.

  • serkanh/cloudwatch-logs-mcp: Facilitates AI-driven access to AWS CloudWatch logs, enabling listing and retrieval of log entries through an MCP server.

  • QAInsights/k6-mcp-server: Facilitates running k6 load tests through a Model Context Protocol server with customizable test parameters and real-time execution output.

  • tay3223/mcp-server-prometheus: Integrate Prometheus monitoring data into the Cursor code editor for real-time insights and historical trends.

  • besthand/mcp-server-taiwan-aqi: Provides real-time and historical air quality data from Taiwan's monitoring stations via an MCP server.

  • dortegau/mcp-proxy-sidecar: Enhances the JetBrains MCP Server with real-time WebSocket monitoring for all MCP interactions.

  • getsentry/sentry-mcp-stdio: Facilitates interaction with Sentry's API for AI-driven error analysis and project management.

  • suolr/lighthouse-integration: Enhances Puppeteer MCP with Google Lighthouse for comprehensive web performance audits and accessibility checks.

  • pab1it0/prometheus-mcp-server: Enables AI assistants to query and analyze Prometheus metrics using standardized interfaces.

  • asatasinon/mcp-datadog-api: Facilitates querying and analyzing Datadog data through an MCP server with integration capabilities for Claude for Desktop.

  • elliottlawson/mcp-sentry: Facilitates AI-driven analysis of Sentry.io issues by providing detailed error reports and debugging information.

  • marcoeg/mcp-server-ntopng: Enables AI agents to query network monitoring data using NTOPNG's database.

  • MaxWeeden/mcp-dnsdumpster: Facilitates AI-driven DNS reconnaissance by interfacing with the DNSDumpster API through natural language requests.

  • weetime/prometheus-mcp-server: A TypeScript-based server bridging Prometheus metrics with Claude via the Model Context Protocol.

  • idoru/influxdb-mcp-server: Facilitates querying and managing InfluxDB instances through an MCP server interface.

  • kaznak/alertmanager-mcp: Integrates Prometheus Alertmanager with Claude AI for seamless alert management and interaction.

  • amitdeshmukh/stdout-mcp-server: Facilitates real-time log capture and analysis from multiple processes via a named pipe system, enhancing debugging and monitoring capabilities.

  • pydantic/logfire-mcp: Logfire MCP Server enables LLMs to access and analyze OpenTelemetry traces and metrics using Logfire APIs, facilitating detailed telemetry data retrieval and SQL query execution.

  • jakenuts/mcp-solarwinds: Access and visualize SolarWinds Observability logs with advanced filtering and customizable time intervals.

  • MCP-100/mcp-sentry: Facilitates the retrieval and analysis of issues from Sentry.io, providing detailed insights into error reports and debugging information.

  • codyde/mcp-sentry-ts: Facilitates AI-driven interaction with Sentry for error analysis, project management, and performance monitoring.

  • 8bitgentleman/activitywatch-mcp-server: Facilitates interaction between ActivityWatch time tracking data and LLMs like Claude, enabling advanced data queries and settings retrieval.

  • winor30/mcp-server-datadog: Facilitates seamless integration with Datadog's monitoring features through an MCP server, enabling efficient incident management and observability.

  • obutora/go_sentry_mcp_server: Facilitates the retrieval of issue lists from Sentry using a Go-based MCP server.

  • ravitemer/mcp-hub: Centralized management and monitoring hub for multiple MCP servers, offering dynamic server lifecycle handling and a REST API for tool execution.

  • orliesaurus/pulsemcp-server: Facilitates the discovery and exploration of MCP servers and integrations via the PulseMCP API.

  • HosakaKeigo/claude-weather-api: A mock API designed for integration with the openapi-mcp-server, enabling weather queries through Claude Desktop.

  • OpsBeaconInc/mcp-opsbeacon-server: Facilitates interaction with the Opsbeacon API through Claude, offering command and connection listing capabilities.

  • Ivlad003/mcp_newrelic: Facilitates querying New Relic logs with NRQL through a Model Context Protocol server, enabling seamless interaction with LLMs like Claude.

  • grafana/mcp-grafana: Facilitates interaction with Grafana's ecosystem, enabling dashboard searches, datasource queries, and incident management through MCP.

  • macrat/mcp-ayd-server: Facilitates status monitoring for Ayd through an MCP server interface.

  • ueki-tomohiro/sentry-issue-server: A TypeScript-based MCP server implementing a notes system with tools for note creation and summarization.

  • chatmcp/mcp-server-collector: Facilitates the discovery and submission of MCP servers by extracting them from URLs and content, and submitting them to directories.

  • loglmhq/mcp-server-prometheus: Facilitates interaction with Prometheus metrics and data through a TypeScript-based MCP server, bridging Claude and Prometheus via the Model Context Protocol.

  • sqr/datadog-mcp: Facilitates log retrieval from Datadog for analysis by Claude within a Kubernetes environment.

  • daedalus/mcp-zabbix - MCP server for Zabbix API.

  • vdineshk/dominion-observatory: Behavioral trust scoring for MCP servers. Continuously probes 22,500+ servers for uptime, latency, and success rate; provides trust_score (0-100) and behavioral drift detection via a free REST API.

  • unitedideas/resolve-mcp: Structured error recovery for AI agents. Returns resolution playbooks with backoff schedules, retry strategies, and recovery steps for 20+ services including OpenAI, Anthropic, Stripe, AWS, and Postgres.

  • smigolsmigol/llmkit: AI cost tracking and budget enforcement MCP server with 11 tools for monitoring spend, session costs, and token usage across Claude Code, Cursor, and Cline.

  • friendlygeorge/system-monitoring-mcp-server: Linux host monitoring MCP server with 19 tools for CPU, memory, disk, network, processes, systemd services, and system logs. Agent-native structured JSON from /proc, /sys, journald. Install: npx @supernova123/system-monitoring-mcp-server. MIT. No API keys required.

  • iPythoning/domain-monitor-mcp-server - Domain WHOIS and SSL certificate monitoring via RDAP and crt.sh — single or batch domain checks with severity classification. Zero API keys, stdio transport (Claude Code/Desktop, Cursor, Windsurf). Install: npx domain-monitor-mcp-server.

  • prismix/status-mcp: Real-time AI service status MCP server tracking 75+ providers (OpenAI, Anthropic, Gemini, Mistral, etc.) with uptime and incident feeds. Streamable HTTP transport (protocol 2024-11-05), no auth required. Claude Desktop config: {"mcpServers":{"prismix":{"url":"https://prismix.dev/api/v1/mcp"}}}. Tools: check_ai_status (query a specific provider), list_ai_services (list all 75+ tracked services).

  • Necmttn/ax: Local-first MCP server over an AI coding-agent telemetry graph: recall sessions, inspect costs, skills, tool usage, and OTLP events across Claude Code, Codex, OpenCode, Cursor, and Pi. Install: curl -fsSL ax.necmttn.com/install | sh; run ax mcp.

  • newrelic-experimental/preflight: Local-first observability MCP server for AI coding assistants — captures every tool call, tracks USD cost per session/day/week and per model, scores efficiency, and detects anti-patterns (re-reads, blind edits, stuck loops) on a live local dashboard. Offline by default (Apache-2.0); optional New Relic backend for team rollups and alerting. Works with Claude Code, Cursor, Windsurf, Copilot, Zed, Continue.dev, and Amazon Q. Install: npx @newrelic/preflight setup.

  • corewebvitals/state-of-cwv-mcp: Query Core Web Vitals and technique metrics across many sites, filter/group by CMS, CDN, and framework, and pull metric histograms. Same data as the public State of Web Vitals explorer.

  • shibley/mymcptools: Uptime and schema-drift monitoring for other MCP servers — every verdict comes from a real MCP initialize handshake against the server's remote endpoint, not self-reported metadata, so an agent can find a server and check whether it has actually been up. Streamable HTTP, stateless, read-only, no auth: https://mymcptools.com/api/mcp. Tools: search_mcp_servers, get_mcp_server, get_server_status, get_server_history, list_server_incidents, list_schema_drift, list_categories, get_catalog_stats. Note: GET returns 405 (stateless serverless, no held-open SSE stream); POST-based clients work.

  • hyperping/mcp-server: Uptime, API, cron and server monitoring via MCP: monitors, outages and their timelines, uptime, response time, MTTR and MTTA, on-call schedules, escalation policies and notification integrations. 26 tools (22 read, 4 write). Remote streamable-http at https://api.hyperping.io/v1/mcp, Bearer token, no install. Registry id com.hyperping/hyperping. Docs: https://hyperping.com/docs/mcp.

  • AIS Revizor - Local-first runtime visibility and AI-managed debug logging for AI coding agents. Replaces blind print() debugging by turning runtime execution into causal chains, letting agents instantly find the exact breakage point in Python and TypeScript apps. Single binary with MCP, CLI, and HTTP interfaces.

  • Apitella: Schema-drift and security monitoring for MCP servers and REST APIs (prompt injection, missing safety hints, exposed credentials), exposed as MCP tools so an agent can check a dependency's health directly. Remote streamable-http: https://api.apitella.io/v1/mcp (API key, 5 tools — list/add sources, get status, recent drift, poll now) or https://api.apitella.io/v1/mcp/public (keyless scan tool, no signup).

  • Spike: Official hosted Spike MCP server for connecting AI assistants to Spike incident management and on-call workflows. Streamable HTTP endpoint at https://mcp.spike.sh/mcp; OAuth authentication; supports Claude, ChatGPT, and Cursor. Hosted service.

  • Continuum-AI-Corp/OrcaReplay: Lets an agent read and re-run its own past sessions. Records a coding-agent run below the harness (model API, shell exit codes and timing, per-turn file changes, MCP calls) and serves that trace store over stdio: orca_list_runs, orca_show_run, orca_checkpoints, orca_graph (what caused what, each edge labelled recorded or inferred), orca_replay (re-runs the recording offline, free and no tokens) and orca_compare (which states in its own description that it spends real tokens). Answers "why did you delete that file yesterday" from the recording rather than from memory. Read-only except orca_compare; replay never opens a network connection. Node 20+, Apache-2.0. Install: npx -y orcareplay mcp.

  • renduples/sitepassport-mcp: Live-checks whether a WordPress site (or any site) is ready to be safely operated by AI agents — llms.txt, AI-crawler robots.txt directives, schema.org markup, and a WebMCP manifest — pulled from the site itself, never self-reported. Free, no API key. Remote streamable-http at https://sitepassport.org/.well-known/mcp.json. Registry id org.sitepassport/check-wordpress-agent-readiness.

  • Uptimely: Manage uptime monitors, incidents and status pages, and read check results, in Uptimely. Hosted remote endpoint: https://app.getuptimely.com/api/mcp (Streamable HTTP, OAuth 2.1 with dynamic client registration, no API key). Official MCP Registry: com.getuptimely/uptimely. Setup for Cursor, VS Code, Cline and Claude Code: https://github.com/DevinoSolutions/mcp-servers/tree/main/uptimely

  • RealUptime: Hosted MCP server for RealUptime: manage uptime monitors, heartbeat checks, incidents and status pages from an AI client. Remote streamable-http at https://mcp.realuptime.io/mcp, API key authentication, no install. Also listed at https://glama.ai/mcp/connectors/io.realuptime/mcp

  • Launch Ready: Deterministic pre-launch website QA for agents: queue a scan of a staging or production URL, then read the Launch Readiness score, blockers and findings. Findings come from fixed rules, not a model, across technical, SEO, analytics, content, forms, performance, accessibility and compliance checks. Tools: scan_site, get_report, list_blockers, list_findings, get_finding, list_projects, create_share_link (Agency plan). Remote Streamable HTTP endpoint: https://uselaunchready.com/api/mcp. Auth: workspace API key (lr_live_…) as a Bearer token, available on every plan including Free. Docs: https://uselaunchready.com/docs/api#mcp-connector. Official MCP Registry: io.github.ochodigits/launch-ready.

  • Tá Rodando: Uptime monitoring and public status pages in Brazilian Portuguese. Agents can create and manage HTTP monitors, alert channels, status-page configuration, and incidents through 19 tools. Remote Streamable HTTP endpoint: https://tarodando.com.br/mcp; OAuth 2.1 with dynamic client registration or an API key as a Bearer token. Free plan available. Official MCP Registry id br.com.tarodando/mcp. Proprietary hosted service.

  • marvtub/nimo-mcp: Website monitoring for your AI assistant: uptime checks every 5 minutes with Slack or Telegram alerts, page speed and Core Web Vitals (real Chrome visitor data when available plus Lighthouse) with a plain-English explanation and the next fix, cron/backup check-ins that alert when a job doesn't run, and scheduled audits with before/after comparison....

  • Parallel-Platforms/hootlens-mcp: Official Hoot Lens server for session replays and click heatmaps from real website visits. Find clicks that broke, see the JavaScript error and source line behind them, and check whether a fix helped. Hosted at https://mcp.hootlens.com/mcp with OAuth, or run locally with npx -y @hootlens/mcp. Demo with sample data, no sign-in: https://mcp.hootlens.com/demo.

  • Blackman99/postsleuth: Answers "why isn't Postfix delivering?" — reads the mail queue (postqueue -j), the mail log and DNS, groups every stuck message by why it is stuck (rate limit, blocklist, dead host, auth, TLS, …) and says what to do next; also traces a message through the log, shows per-destination deferral trends, and checks SPF/DKIM/DMARC/PTR. Read-only MCP (postsleuth mcp: postfix_doctor, postfix_queue_summary, postfix_queue_shape, postfix_log_trace, postfix_log_trend, …); requeue/flush/hold/delete appear only with --allow-mutations and each call needs confirm: true. Install: go install github.com/Blackman99/postsleuth/cmd/postsleuth@latest or a GoReleaser binary. Transport: stdio. Auth: none. License: MIT.

  • LogneBudo/llmxray: Lets an agent ask what went wrong with its own Ollama calls, in sentences with the fix: lost prompt cache and the token where the prompt stopped matching, context overflow, a model running partly on CPU, tool schemas crowding the prompt; plus the last turn in numbers and which local models fit in video memory. Local, read-only, Apache-2.0. Install: npx -y -p llmxray llmxray-mcp.

  • Watchbot MCP server: Find business leads, monitor web pages and read change diffs, fetch pages as text or structured prices, and track grant and tender keywords. Hosted Streamable HTTP at https://mcp.watchtbot.site/mcp; pass a Watchbot API key as the api_key tool argument. Prepaid pay-per-call service with free signup credits; proprietary hosted service.

  • 01men/ybkkasi: Puts MCP servers, skills, agents and AI apps into a single self-hosted asset ledger so an agent can answer "who owns this, what is it allowed to do, and which team is paying for it". Lifecycle state machine per asset (draft → active → deprecated → retired), 84 per-tool RBAC permission points with declared minimum scopes, append-only audit trail, dependency topology for change-impact analysis, and cost attribution by department × agent × skill. Ships 27 plugin packages, 354 REST routes and 83 MCP tools; npm test runs 1467 end-to-end assertions. Transport: Streamable HTTP at POST /mcp (JSON-RPC 2.0). Auth: platform Bearer token — the MCP endpoint requires authentication by design, so anonymous introspection will fail. Install: git clone https://github.com/01men/ybkkasi && npm ci && ADMIN_PASSWORD='<strong-password>' npm start (Node 22+, no build step; a Dockerfile and docker-compose.yml are included but the published container image is not up yet). Read-only demo instance with seeded data: https://demo.ybkkasi.com. License: MIT.

  • elfovo/agent-chain-doctor: Read-only diagnostic for AI agents scheduled with launchd, cron or systemd (e.g. claude -p overnight): reads the traces the chain leaves on disk and runs 30 checks for the ways an unattended agent stops silently (a scheduler PATH without the agent binary, a stale lock no one clears, a truncated wake-up file, a laptop that sleeps through the window...), each answered exposed, guarded or undecidable with its evidence. One tool, diagnose_scheduled_agent. Install: git clone https://github.com/elfovo/agent-chain-doctor && claude mcp add agent-chain-doctor -- python3 "$PWD/agent-chain-doctor/mcp/server.py". Transport: stdio. Auth: none. Python 3 standard library only, no network, no writes. Registry id io.github.elfovo/agent-chain-doctor-mcp. MIT.

  • Zabbix MCP Server by initMAX: Query and manage Zabbix hosts, items, triggers, problems, templates and dashboards through 237 MCP tools; investigate incidents, render metric graphs, detect anomalies, forecast capacity and generate PDF reports. Self-hosted, multi-server, with OAuth 2.1 and read-only access controls.

  • becoolme/openuptime: Manage uptime monitors, incidents and public status pages. Remote Streamable HTTP at https://openuptime.app/api/mcp; OAuth authorization required, API bearer token supported. Install: claude mcp add --transport http openuptime https://openuptime.app/api/mcp. Docs: https://openuptime.app/docs/mcp. Registry id app.openuptime/mcp. TypeScript, MIT.

  • MCP Checkup: Check public remote MCP availability, protocol handshakes, tool and schema drift, and authentication metadata. Results are published as verifiable Ed25519-signed DSSE attestations. Tools: check_mcps, get_mcp_report, get_mcp_history, get_attestation, search_mcps, request_check (contacts a third-party server). Hosted Streamable HTTP at https://mcpcheckup.com/mcp; no auth. Official MCP Registry: com.mcpcheckup/mcp. Hosted service; check suite and verifier: Apache-2.0.

  • NIGHTSHIFT (public): Discover specialized agent crews and try free samples, including evidence-backed monitoring of Ethereum tokens and Uniswap v4 hooks. Start with start_here. Hosted Streamable HTTP at https://getnightshift.tech/mcp/public; public discovery requires no key. Source license: MIT. Docs.

  • API Meter: Per-customer API usage analytics for API owners and their agents: create an account, get server ingest snippets (Express, Hono, FastAPI, Flask, Django, Go, curl), and read or export monthly usage by customer, endpoint and plan, with AI-agent traffic in its own bucket. Tools: create_account, get_ingest_snippet, get_usage, export_usage. Remote streamable-http at https://api-meter-lab.pages.dev/mcp; no auth for signup and snippets, dashboard bearer key (msd_...) for usage reads. Registry id io.github.Surface-ops/api-meter. License: MIT.

  • Staygreen: Uptime and incident monitoring for apps built with coding agents: read status, monitors, uptime, response times, incidents, and SSL and domain expiry; scan a public app URL; set up an app's monitors and status page in one call; pause monitors; draft incident updates. Hosted Streamable HTTP at https://staygreen.io/mcp; OAuth 2.1 or a workspace API token.

  • UserProof (io.github.GetUserProof/userproof): Official server for UserProof, UX analytics for coding agents: it records real visitor sessions on your site and returns the one highest-impact conversion fix (dead clicks, rage clicks, errors on click, drop-offs) with the sessions behind it and the file and line to change, then checks after you ship whether the fix worked. 13 tools: get_top_fix, check_fix (before and after a shipped fix), mark_fix_applied, set_goal, scan_site (pre-traffic scan of the live site, no form posts), check_coverage (pages missing the snippet), wait_for_first_event, fleet, list_sites, select_site, and account tools get_usage, get_activation, upgrade. Local stdio, one dependency-free file (Node 18+): npx -y github:GetUserProof/userproof; auth: a UserProof API key created by the setup flow (~/.config/userproof/key). Also a Claude Code plugin with setup, review and go skills: /plugin marketplace add GetUserProof/userproof. Official MCP Registry: io.github.GetUserProof/userproof. License: MIT.

  • WatchCron: Manage cron-job monitors, uptime checks and incidents, including schedules, ping history, pause/resume and notification channels. Local stdio. Install: npx -y watchcron-mcp; requires WATCHCRON_API_KEY from https://watchcron.com. Clients: Claude Desktop, Claude Code, Cursor, Windsurf. MIT.

  • SaaS Pro Max: Inspect product analytics, errors, alerts and feature flags in one selected organization. Users choose Read, Read and change, or Full OAuth access, with server-side organization permissions still enforced. Broader grants allow app, flag, people and alert-rule changes; ingestion-key operations require Full access. Endpoint: https://saaspro.dev/api/mcp; transport: streamable-http; auth: oauth. License: Private hosted-service implementation; no public server-source license claim..

  • Kilobat CLI: Inspect macOS network conditions and diagnose Wi-Fi or host-connection problems. Install: brew install seifip/kilobat/kilobat; start: kilobat mcp. Local stdio; no auth. GPL-3.0-only; the desktop app is proprietary.