An Amplitude MCP that reads the whole analytics surface and deletes nothing
Connect one API key and secret key pair and real-time users, event segmentation, funnels, retention, revenue LTV, per-user activity streams, chart annotations, cohorts, and server-side event tracking become workflow steps for visual automations and AI agents. All 23 nodes are reads or additive writes: nothing in this connector can delete or damage analytics data.
What the Amplitude MCP connector does
The Amplitude connector for Mattering Automations is 23 nodes across Amplitude's server-side APIs: real-time active users in five-minute buckets, daily, weekly, and monthly active and new user counts, user composition by property, session length and sessions-per-user metrics, event segmentation with metrics, formulas, and group-bys, funnel and retention analyses, revenue lifetime value cohorts, user search with per-user event streams, chart annotations, taxonomy reads for schema governance, behavioral cohort listing, upload, and membership management, and event, user property, and group property ingestion.
Every node speaks MCP, the open protocol for connecting AI assistants to tools, so Claude, ChatGPT, and any MCP-capable assistant can use them, alongside Matter, the AI assistant built into the platform. An assistant can answer how many users are active right now, where the activation funnel leaks, and how last month's cohorts retained, without anyone opening the Amplitude dashboard.
Because Mattering is a work platform rather than middleware, analytics reads land somewhere useful by default: a weekly retention and funnel digest can arrive as a report task in Mattering Productivity, a churn-risk cohort update can drive follow-up tasks for the team, and query results can update records in a Mattering database, all on one canvas next to the rest of the work.
The safety posture is worth stating plainly: 17 of the 23 nodes are pure reads, and the six writes create chart annotations, send events and properties into Amplitude, or manage cohort membership. Nothing in the catalog deletes or destroys anything, so no workflow built from these nodes can damage the analytics data your team depends on.
One connection, three surfaces
- Visual automations. Drag Amplitude nodes onto the canvas and wire them to triggers, conditions, and other apps.
- AI agents, yours or ours. These nodes speak MCP, the open protocol, so Claude, ChatGPT, and any MCP-capable assistant can use them, and Matter, the agent built into the platform, uses them natively.
- The whole platform. Combine with native tasks, projects, and other connectors, no middleware in between.
Every Amplitude node, ready to drop on the canvas
23 nodes, generated from the connector itself so this list always matches what ships. Expand a category to browse.
Read and query nodes 17
Amplitude: Get Active or New User Counts
Get daily, weekly, or monthly active or new user counts over a date range, optionally segmented and grouped by a property: the DAU/WAU/MAU read. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Get Average Session Length
Get the average session length in seconds for each day in a date range. Session lengths are capped at one day. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Get Average Sessions per User
Get the average number of sessions per user for each day in a date range: an engagement-depth companion to the DAU read. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Get Real-Time Active Users
Get active user counts in five-minute buckets for today and yesterday: the live pulse of the product. Takes no required parameters, so it doubles as the credential health check for the whole connector. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Get Revenue LTV
Get revenue lifetime value cohorts: for users first seen on each date, cumulative revenue per user (ARPU), per paying user (ARPPU), total revenue, or paying-user counts at 1, 7, 14, 30, 60, and 90 days out. Only returns data if the project tracks revenue. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Get User Composition by Property
Break the active user base down by one property's values over a date range, e.g. how many users per country, platform, app version, or plan. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Get a User's Event Stream
Get one user's profile summary (first/last seen, event count, user properties) and their recent event stream by Amplitude ID (find it with the Search for a User node). Event streams are personal data and persist in execution records; keep limit modest. Budget: 360 queries per hour, 10 concurrent (shared with user search).
Amplitude: List Chart Annotations
List the project's chart annotations (release markers, campaign launches, incidents), optionally filtered by category, chart, or date window. Each annotation carries its id, label, timestamps, and category.
Amplitude: List Cohorts
List the project's behavioral cohorts with id, name, size, definition metadata, and timestamps. Note: downloading a cohort's MEMBER LIST is a separate async export job this connector deliberately does not model (the file can be millions of rows); manage membership with the upload and membership nodes instead. Cohort API budget: 60 requests per 10 minutes.
Amplitude: List Event Properties (Taxonomy)
Read the planned properties of one event type (or the shared properties when no event type is given): names, value types, enum values, and required flags. Plan gate: requires an Amplitude Enterprise plan.
Amplitude: List Event Types (Taxonomy)
Read the project's planned event taxonomy: every event type with its category, description, and visibility flags, for schema governance and discovery. Plan gate: the Taxonomy API requires an Amplitude Enterprise plan; other plans get a permission error, not an auth failure.
Amplitude: List Tracked Events
List every event tracked in the project with its display name, weekly totals, and visibility flags: the schema discovery step before building segmentation or funnel queries. Hidden events are not returned. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: List User Properties (Taxonomy)
Read the project's user property dictionary: standard properties (device_id, user_id, amplitude_id, country and friends) plus custom properties (prefixed gp:), with types and descriptions. Plan gate: requires an Amplitude Enterprise plan.
Amplitude: Run a Funnel Analysis
Compute funnel conversion across an ordered list of events: step-by-step and cumulative conversion rates, raw counts, and median/average transition times between steps. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Run a Retention Analysis
Get cohort retention: for users who did a start event, how many came back and did a return event in each subsequent period. Supports n-day (the default), rolling, and bracket retention modes. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Amplitude: Search for a User
Find a user's Amplitude ID by searching with an Amplitude ID, a device ID, a user ID, or a user ID prefix. Feed the returned amplitude_id into the Get User Activity node. Budget: 360 queries per hour, 10 concurrent (shared with user activity).
Amplitude: Segment an Event
The workhorse analytics read: count one event (optionally compared against a second) over time with a chosen metric (uniques, totals, pct_dau, average, property sums or averages, or a formula), filtered by segments and grouped by up to two properties. Dashboard REST budget: cost-based, 108,000 cost per hour and at most 5 concurrent requests shared across all Amplitude query nodes; cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Action nodes 6
Amplitude: Add or Remove Cohort Members
Incrementally ADD or REMOVE members of an existing uploaded cohort without re-uploading the whole list. Reversible by construction: every ADD has a matching REMOVE. Cohort API budget: 60 requests per 10 minutes.
Amplitude: Create a Chart Annotation
Add a chart annotation, e.g. mark a deploy, a campaign start, or an incident so it shows on Amplitude charts at that timestamp. Reversible: annotations can be deleted in the Amplitude UI or via the annotations API.
Amplitude: Set Group Properties (Group Identify)
Set properties on a GROUP (e.g. an organization or account) rather than a user, keyed by group type and value. Group analytics features require an Amplitude plan with Accounts enabled. $set overwrites existing values for the named properties. Auth note: this endpoint takes only the project API key (sent in the request body); the runner injects it from the connector credential's username, and the secret key is never sent to the ingestion host.
Amplitude: Set User Properties (Identify)
Update a user's properties without sending an event, using Amplitude's Identify operations ($set, $setOnce, $add, $append, $unset and friends). Careful: $set OVERWRITES the existing value of each property it names; changes apply to future events only, never retroactively. Budget: 1,800 property updates per hour per user. Auth note: this endpoint takes only the project API key (sent in the request body); the runner injects it from the connector credential's username, and the secret key is never sent to the ingestion host.
Amplitude: Track Events (HTTP V2)
Send events INTO Amplitude from an automation (server-side tracking). Each event needs an event_type plus a user_id or device_id (5+ characters); include time (epoch ms) and an insert_id so retries deduplicate. Up to 2,000 events per request, 1 MB max payload. Note: ingested events cannot be deleted through this connector, so point test runs at a throwaway project or a clearly named test event. Auth note: this endpoint takes only the project API key (sent in the request body); the runner injects it from the connector credential's username, and the secret key is never sent to the ingestion host.
Amplitude: Upload a Cohort
Create (or replace, via existing_cohort_id) a static cohort from a list of user identifiers, e.g. sync a customer list from another system into Amplitude for targeting and analysis. Up to 100,000 identifiers per request. Reversible: an uploaded cohort can be deleted in the Amplitude UI. Cohort API budget: 60 requests per 10 minutes.
Connecting Amplitude takes one credential
Every Amplitude project carries an API key and a secret key, shown in the project's settings (Settings, then Projects, then your project). The connector stores them as a single HTTP Basic credential: the API key is the username and the secret key is the password. The same pair covers the entire connector; the three ingestion nodes need only the API key half, and the connector sends just that half for them automatically, so there is nothing extra to configure.
In Amplitude, open Settings, go to Projects, select the project you want to automate, and copy its API key and secret key.
In Mattering Automations, open Connectors, click Add Connector, choose Amplitude, and create a credential with the API key as the username and the secret key as the password.
Click Test Connection. Mattering calls Amplitude's real-time active users endpoint, which works on every Amplitude plan and is one of the cheapest reads on the query budget.
All 23 Amplitude nodes appear in the canvas palette. Run Amplitude: List Tracked Events first: it lists every event the project tracks, which is the schema discovery every segmentation, funnel, and retention query builds on.
Common uses for the Amplitude MCP
The live pulse and the daily actives
Amplitude: Get Real-Time Active Users returns active user counts in five-minute buckets for today and yesterday, and Amplitude: Get Active or New User Counts covers the daily, weekly, and monthly active and new user reads over any date range. Amplitude: Get User Composition by Property breaks the user base down by country, platform, app version, or plan, and Amplitude: Get Average Session Length with Amplitude: Get Average Sessions per User add the engagement-depth reads.
Segmentation with metrics and formulas
Amplitude: Segment an Event is the workhorse read: count one event over time, optionally against a second, with a chosen metric such as uniques, totals, percent of daily actives, property sums or averages, or a formula, filtered by segments and grouped by up to two properties. Amplitude: List Tracked Events maps what exists first, and on Enterprise plans the taxonomy nodes read the planned schema itself so an assistant can learn what fields mean before querying.
Funnels, retention, and revenue
Amplitude: Run a Funnel Analysis computes step-by-step and cumulative conversion across an ordered list of events with median and average transition times, and Amplitude: Run a Retention Analysis reports how users who did a start event came back to do a return event, with n-day, rolling, and bracket modes. Amplitude: Get Revenue LTV adds cumulative revenue per user at 1 through 90 days out for each acquisition cohort, in projects that track revenue.
One user's whole story
Amplitude: Search for a User translates an Amplitude ID, device ID, user ID, or user ID prefix into the numeric Amplitude ID, and Amplitude: Get a User's Event Stream returns that person's profile summary and recent events. Event streams are personal data and persist in execution records, so the node keeps its limit modest by default; pull only what the investigation needs.
Release markers on every chart
Amplitude: Create a Chart Annotation stamps a deploy, campaign launch, or incident onto Amplitude's charts at that timestamp, and Amplitude: List Chart Annotations reads the markers back, filtered by category, chart, or date window. Wire the create node into the end of a release workflow and the analytics team sees exactly when each version shipped, right on the charts they already watch.
Cohorts in both directions, and events in
Amplitude: List Cohorts enumerates the project's behavioral cohorts with sizes, Amplitude: Upload a Cohort syncs a customer list from another system into Amplitude with up to 100,000 identifiers per request, and Amplitude: Add or Remove Cohort Members adjusts membership incrementally, reversibly by construction. On the ingestion side, Amplitude: Track Events (HTTP V2) sends events from automations, and Amplitude: Set User Properties (Identify) with Amplitude: Set Group Properties (Group Identify) update user and group properties without sending an event.
Example: A weekly retention and funnel digest that lands as a task, with drop-offs opening their own
Built on the Automations canvas in a few minutes, or described in plain language to Matter and built for you.
Trigger: a schedule fires Monday morning.
Amplitude: Get Active or New User Counts pulls the week's daily active and new user numbers.
Amplitude: Run a Funnel Analysis computes the activation funnel's step-by-step conversion for the week.
Amplitude: Run a Retention Analysis reports how the last month's signup cohorts came back.
A condition compares each funnel step's conversion to the prior week and opens an investigation task in Mattering Productivity for any step that dropped.
The digest itself, actives, new users, funnel conversion, and retention, lands as a report task in Mattering Productivity, and the headline numbers update a record in a Mattering database for the long-run trend.
Amplitude MCP questions, answered
What Amplitude operations are supported?
23 nodes: 17 queries and 6 actions. The queries cover real-time active users, active and new user counts, user composition, session length and sessions per user, tracked event listing, event segmentation with metrics and formulas, funnels, retention, revenue LTV, user search and per-user event streams, chart annotation listing, cohort listing, and the three taxonomy reads. The actions create chart annotations, upload cohorts, add or remove cohort members, track events, set user properties, and set group properties. Deliberately not shipped: the raw Export API (see the dedicated question below), the cohort member-list download job, the bulk batch ingestion endpoint, taxonomy writes, annotation update and delete, chart CSV export, and Amplitude's admin, attribution, and compliance surfaces.
Does this work with Claude, ChatGPT, or other AI assistants?
Yes. Mattering exposes these nodes over MCP, an open protocol for connecting AI assistants to tools, so any MCP-capable assistant, including Claude and ChatGPT, can connect to your workspace and use them. Matter, the assistant built into the platform, uses the same nodes natively. With 17 of the 23 nodes being pure reads and nothing destructive anywhere, this is an unusually safe connector to hand an agent.
Can this connector damage my Amplitude data?
No. There are zero destructive nodes: 17 of the 23 are pure queries, and the six writes only add or update. Honest caveats within that: events sent with Amplitude: Track Events (HTTP V2) cannot be deleted through this connector, so point test runs at a throwaway project or a clearly named test event; the set operation in Amplitude: Set User Properties (Identify) and Amplitude: Set Group Properties (Group Identify) overwrites the existing values of the properties it names; and Amplitude: Upload a Cohort pointed at an existing cohort id replaces that cohort's membership. Amplitude: Add or Remove Cohort Members is reversible by construction, since every add has a matching remove, and annotations can be deleted in the Amplitude UI.
Why is raw event export not included?
Deliberately. Amplitude's Export API returns a zipped archive of event files with no row limit parameter, only an hour-range window, and a single request can return up to 4 GB. An archive like that cannot flow through workflow outputs honestly, so the connector omits the endpoint rather than shipping it broken. For the query cases, Amplitude: Segment an Event covers aggregate questions and Amplitude: Get a User's Event Stream covers bounded per-user event pulls; bulk data pipelines belong in Amplitude's own export destinations, outside an automation canvas.
Which nodes depend on my Amplitude plan or project setup?
Three groups. The taxonomy reads, Amplitude: List Event Types (Taxonomy), Amplitude: List Event Properties (Taxonomy), and Amplitude: List User Properties (Taxonomy), require an Amplitude Enterprise plan: on other plans they return a permission error, which is a plan problem, not a credential problem. Amplitude: Set Group Properties (Group Identify) needs a plan with Amplitude's Accounts feature enabled. And Amplitude: Get Revenue LTV only returns data if the project tracks revenue. Everything else works on every plan, including the free one.
How does one credential cover both analytics queries and event ingestion?
The credential is the project's API key and secret key stored as an HTTP Basic pair, and the query, annotation, taxonomy, and cohort nodes authenticate with both halves. Amplitude's ingestion endpoints are different: they take only the API key, sent in the request body. The connector handles this itself, injecting just the API key half for Amplitude: Track Events (HTTP V2), Amplitude: Set User Properties (Identify), and Amplitude: Set Group Properties (Group Identify), so the secret key is never sent to the ingestion host and there is nothing extra to configure.
Can an Amplitude event trigger a workflow?
Not in this connector: there is no inbound Amplitude trigger. The honest pattern is a scheduled sweep: run a query node such as Amplitude: Segment an Event or Amplitude: Get Active or New User Counts on a schedule, compare against a threshold or the prior period, and route what changed into tasks or messages. Mind the shared budget when scheduling: dashboard queries share a cost-based budget of 108,000 cost per hour with at most 5 concurrent requests, and cost grows with the date range, filter conditions, and group-bys, so keep ranges tight.
Does this work with EU data residency?
Not in this version. Amplitude operates a separate EU data center, and this connector deliberately speaks only to the US hosts, so a project stored in the EU will fail to authenticate or return empty data. If your project lives on US residency, everything works as described. EU host support is a candidate for a future version.
Put Amplitude to work inside your workspace
Connect once, then use every node in visual automations and AI agent workflows across the Mattering platform.
All product names, logos, and brands are property of their respective owners and are used for identification purposes only. Use of these names, logos, and brands does not imply endorsement or affiliation. Amplitude is a product of its respective owner; Mattering is an independent platform.