Hosting

remote

Transport

Not documented

Endpoint

https://mcp.amplitude.com/mcp

Authentication

OAuth 2.0 browser flow. Use the EU endpoint https://mcp.eu.amplitude.com/mcp when the organization's data resides in the EU. Access is restricted to the authenticated account and organization/project roles; organization admins can disable MCP access and RBAC actions control read/write.

Requirements

• Amplitude account with access to the target organization and project
• MCP client with remote Streamable HTTP and OAuth support
• Organization MCP access enabled; project-level role permissions for requested reads/writes

Setup steps

1. For US data, add https://mcp.amplitude.com/mcp as a remote MCP server; for EU residency, use https://mcp.eu.amplitude.com/mcp.
2. Complete the OAuth flow in the browser using the correct Amplitude organization. If multiple organizations are signed in, sign out and authenticate to the intended one.
3. Call get_amplitude_context with a project ID to load org/project AI Context; use progressive discovery by adding ?discovery=progressive when supported to lower initial tool context.
4. For Codex CLI, the official command is codex mcp add amplitude --url https://mcp.amplitude.com/mcp, then complete OAuth.
5. Verify with a read-only prompt such as ?What Amplitude projects can I access?? before using edit or delete actions. Review and confirm every proposed change.

Amplitude’s current docs describe a full default catalog, an approximately 25-tool ChatGPT subset, and a separate progressive-discovery mode. The stored 31 names are an account/client-dependent sample rather than a complete universal list; retain partial_dynamic and document the discovery flow. Analytics tools can write content and taxonomy; RBAC gates writes.

• get_amplitude_context
  01
• query_amplitude_data
  02
• render_amplitude_chart
  03
• get_from_url
  04
• use_amp_dashboards
  05
• get_amplitude_charts
  06
• get_experiments
  07
• query_experiment
  08
• get_session_replays
  09
• list_session_replays
  10
• get_session_replay_events
  11
• create_experiment
  12
• create_flags
  13
• create_metric
  14
• update_flag
  15
• use_amplitude_cohorts
  16
• get_amp_taxonomy
  17
• get_transformations
  18
• get_group_types
  19
• manage_amp_events
  20
• manage_amp_properties
  21
• manage_amp_taxonomy
  22
• get_data_ingestion_sources
  23
• get_data_source_details
  24
• get_data_warehouse_destinations
  25
• get_data_warehouse_jobs
  26
• list_tool_categories
  27
• get_category_tools
  28
• describe_tool
  29
• list_guides_surveys
  30
• get_guide_or_survey
  31

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

• This authenticated analytics MCP is distinct from the public read-only Amplitude Docs MCP at https://amplitude.com/docs/api/mcp.
• It is not an event ingestion endpoint; use Amplitude SDKs or HTTP V2 API to send production events.
• Write operations require project/organization permissions; taxonomy deletes require explicit confirmation, and some planned/custom items cannot be restored.
• ChatGPT gets an approximately 25-tool curated subset and lacks session replay/progressive discovery tools.
• Large chart results may be truncated; saved dashboard filters may not apply to chart queries from dashboards.
• Amplitude organization administrators can disable MCP; role-based read/write permission applies per project.
• No runtime connection test performed.

• official tool reference

  Amplitude’s current docs describe a full default catalog, an approximately 25-tool ChatGPT subset, and a separate progressive-discovery mode. The stored 31 names are an account/client-dependent sample rather than a complete universal list; retain partial_dynamic and document the discovery flow. Analytics tools can write content and taxonomy; RBAC gates writes.

• official documentation or source

  Amplitude’s current docs describe a full default catalog, an approximately 25-tool ChatGPT subset, and a separate progressive-discovery mode. The stored 31 names are an account/client-dependent sample rather than a complete universal list; retain partial_dynamic and document the discovery flow. Analytics tools can write content and taxonomy; RBAC gates writes.

Documentation checked 2026-10-28T14:15:00.000Z. Connection has not been runtime tested.