Skip to main content
hc
Submit
3 compared

Product analytics MCP servers compared

Product-analytics MCP servers that bring usage data into Claude, compared on trust, setup, and safety.

Open in the interactive comparison tool
FieldAmplitude MCP Server for Claude

Official remote MCP server for querying and creating Amplitude analytics, dashboards, experiments, cohorts, session replay insights, and product data from Claude and other MCP clients.

Open dossier 		PostHog MCP Server for Claude

Connect Claude to PostHog's official MCP server for product analytics, feature flags, experiments, dashboards, error tracking, logs, surveys, data warehouse, and project operations.

Open dossier 		Google Analytics MCP Server

Experimental Google Analytics MCP server that lets Claude retrieve GA4 account and property details, Google Ads links, custom dimensions and metrics, core reports, funnel reports, and realtime reports.

Open dossier
Trust
Install riskReview firstReview firstReview first
Notes Safety ✓ Privacy ✓ Safety ✓ Privacy ✓ Safety ✓ Privacy ✓
Categorymcpmcpmcp
Sourcesource-backedsource-backedsource-backed
AuthorAmplitudePostHogGoogle Analytics
Added2026-06-052026-06-052026-06-06
Platforms
Claude CodeClaude Desktop
Claude CodeClaude Desktop
Claude CodeClaude Desktop
Source repo———
Safety notes✓The authenticated Amplitude MCP server can expose tools that create and edit analytics content, not just read existing dashboards. Use the existing Amplitude permission model and keep users scoped to the projects and organizations they are allowed to inspect. Require human confirmation before Claude creates dashboards, edits notebooks, changes experiments, creates cohorts, updates metrics, or touches feature-flag workflows. Use the EU MCP endpoint only when the Amplitude data resides in the EU region; otherwise use the standard US endpoint. Consider progressive discovery to reduce tool-list size and force the assistant to describe a tool schema before calling higher-impact tools. Organization administrators can restrict or disable MCP access in Amplitude content access settings; use that control if AI access is not approved.✓The hosted PostHog MCP endpoint is authenticated with a personal API key. Treat that key as a production credential and grant only the read or write scopes needed for the current task. The example configuration uses PostHog's documented `features` query parameter to reduce the exposed tool surface. Feature and tool filtering helps the model choose fewer tools, but API-key scopes remain the real authorization boundary. If no features or tools are specified, PostHog documents that all available tools are exposed. Review the tool list before using an unrestricted endpoint in a production project. Some PostHog MCP tools are mutating or destructive, including tools for creating, updating, archiving, deleting, publishing, enabling, or running product configuration. Keep MCP client approval prompts enabled for those calls. Workflows, alerts, feature flags, experiments, surveys, CDP functions, and data warehouse changes can affect real users, messages, experiments, dashboards, or operational pipelines. Use a development project first. SQL, analytics, error-tracking, log, person, cohort, replay, and data warehouse queries can return large or sensitive result sets. Limit date ranges, filters, and row counts before approving broad queries. Use `https://mcp-eu.posthog.com/mcp` for EU PostHog projects when you need OAuth flows to route to the EU PostHog instance, as documented by PostHog.✓Google Analytics MCP Server is marked experimental upstream, so verify tool behavior and API coverage before relying on it for operational reporting. The server uses Google Analytics Admin and Data APIs through the permissions granted to the configured credentials. Report prompts can generate expensive or broad API queries if date ranges, dimensions, metrics, filters, or funnels are left open-ended. Analytics results can influence marketing, product, or revenue decisions; validate important findings in Google Analytics or internal BI tools before acting. The server is read-oriented, but it still exposes business-sensitive analytics data to the MCP client and model context.
Privacy notes✓Amplitude MCP can expose product analytics, event data, user cohorts, experiment results, dashboard definitions, session replay context, feedback summaries, AI agent analytics, and account health signals to the connected AI client. AI providers may process the Amplitude data returned through MCP; review internal AI, privacy, GDPR, CCPA, and customer-data policies before connecting production projects. Assistant transcripts, MCP client logs, screenshots, support tickets, and exported chats can retain user identifiers, cohort definitions, behavioral data, experiment details, and roadmap-sensitive product insights. OAuth ties access to the current Amplitude user, but it does not make downstream model transcripts private by itself.✓PostHog MCP can return project metadata, event names, event properties, person and group identifiers, cohorts, feature flag rules, experiment results, dashboards, recordings metadata, errors, logs, warehouse query results, prompts, and LLM analytics data into the model conversation. The MCP server acts as a proxy to your PostHog instance. PostHog documents that queries execute against the connected project and results are returned directly to the AI client, with active project or organization session state cached temporarily using an API-key hash. The hosted server runs on Cloudflare infrastructure. Review PostHog's data processing note, your PostHog region, and your own data-processing agreements before sending regulated customer analytics through an AI client. Do not paste personal API keys, project secrets, raw session recordings, customer identifiers, email addresses, support transcripts, payment data, or exported warehouse rows into prompts or PR comments. Claude transcripts, MCP client logs, screenshots, issue comments, and generated runbooks can retain sensitive PostHog query results outside the access controls and retention settings of the original PostHog project.✓Google Application Default Credentials, OAuth client files, service account impersonation details, project IDs, and local credential paths are sensitive and should not be committed or pasted into prompts. Account names, property IDs, Google Ads links, event names, custom dimensions, custom metrics, audiences, funnel steps, realtime activity, and report outputs can reveal business performance and user behavior. GA4 reports can include location, device, campaign, source, medium, conversion, revenue, or audience information that may be regulated or contractually restricted. Tool responses may be retained by MCP clients, model providers, logs, screenshots, and chat transcripts outside Google Analytics retention controls.
Prerequisites
  • Existing Amplitude customer account with access to the projects you want Claude to analyze.
  • MCP-capable client that supports remote Streamable HTTP servers, such as Claude Code, Cursor, ChatGPT, Codex CLI, Lovable, Kiro, or another compatible client.
  • OAuth authorization through Amplitude for the user and organization being queried.
  • Organization-level MCP access left enabled by an Amplitude administrator.
  • PostHog Cloud or self-hosted PostHog project that you are allowed to connect to an MCP client.
  • Personal PostHog API key created with the MCP Server preset or an equivalent least-privilege scope set.
  • Node.js and `npx` available to the MCP client for the PostHog wizard or `mcp-remote`.
  • MCP-capable client such as Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed, or another Streamable HTTP-compatible client.
  • Python 3.10 or newer with pipx available.
  • Google Cloud project with Google Analytics Admin API and Google Analytics Data API enabled.
  • Google Application Default Credentials for a user or service account with access to the intended Google Analytics accounts or properties.
  • OAuth consent and credential setup that includes the `https://www.googleapis.com/auth/analytics.readonly` scope.
Install
claude mcp add -t http -s user amplitude https://mcp.amplitude.com/mcp
npx @posthog/wizard@latest mcp add
pipx run analytics-mcp
Config
{
  "mcpServers": {
    "amplitude": {
      "url": "https://mcp.amplitude.com/mcp"
    }
  }
}
{
  "mcpServers": {
    "posthog": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@latest",
        "https://mcp.posthog.com/mcp?features=docs,insights,error_tracking,events,data_schema",
        "--header",
        "Authorization:${POSTHOG_AUTH_HEADER}"
      ],
      "env": {
        "POSTHOG_AUTH_HEADER": "Bearer phx_your_personal_api_key"
      }
    }
  }
}
{
  "mcpServers": {
    "analytics-mcp": {
      "command": "pipx",
      "args": ["run", "analytics-mcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "PATH_TO_CREDENTIALS_JSON",
        "GOOGLE_PROJECT_ID": "YOUR_PROJECT_ID"
      }
    }
  }
}
Citations
ClaimUnclaimedUnclaimedUnclaimed
More comparisons, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.