MULTI-APP TOOLKIT
Build with Amplitude + Datadog
- Multi-app
- One MCP session
- 24 actions
- 1 triggers
- Managed auth
MCP
One MCP session, both toolsets
One session gives your product or agent every Amplitude and Datadog tool at once — connect once, list tools, and call them like any single-app session.
// accessToken: mint a short-lived token with the Connect SDK — see the MCP guide
const transport = new StreamableHTTPClientTransport(
new URL("https://remote.mcp.pipedream.net/v3"),
{
requestInit: {
headers: {
Authorization: `Bearer ${accessToken}`,
"x-pd-project-id": "{project_id}",
"x-pd-environment": "production",
"x-pd-external-user-id": "{external_user_id}", // any stable ID for this user in your system
"x-pd-app-slug": "amplitude,datadog",
},
},
},
)
const mcp = new Client({ name: "my-agent", version: "1.0.0" })
await mcp.connect(transport)
const { tools } = await mcp.listTools()
// One list, both toolsets: Amplitude and Datadog tools arrive
// together, each keyed by its own app's slug.
// e.g. run Create Cohort:
const result = await mcp.callTool({
name: "amplitude-create-cohort",
arguments: {
cohortName: "Name",
appId: 10,
},
})import { PipedreamClient } from "@pipedream/sdk"
const pd = new PipedreamClient({
projectId: process.env.PIPEDREAM_PROJECT_ID!,
clientId: process.env.PIPEDREAM_CLIENT_ID!,
clientSecret: process.env.PIPEDREAM_CLIENT_SECRET!,
projectEnvironment: "production",
})
const externalUserId = "{external_user_id}" // any stable ID for this user in your system
const [amplitudeTools, datadogTools] =
await Promise.all([
pd.components.list({ app: "amplitude" }),
pd.components.list({ app: "datadog" }),
])
// One external user owns both connected accounts, so either app's tools
// run on their behalf with the same externalUserId.curl "https://api.pipedream.com/v1/connect/{project_id}/components?app=amplitude" \
-H "X-PD-Environment: production" \
-H "Authorization: Bearer {access_token}"
curl "https://api.pipedream.com/v1/connect/{project_id}/components?app=datadog" \
-H "X-PD-Environment: production" \
-H "Authorization: Bearer {access_token}"
# Connect both accounts to the same external_user_id, then
# configure and invoke either app's tools on that user's behalf.ARCHITECTURE
One user, two connected accounts
Your user connects each account once, under whatever ID they already have in your product. The two stay independent — either can be revoked on its own — and your code reaches both through that one user.
Your product
external_user_id
{external_user_id}Pipedream Connect
Managed identity
Auth · tools · routing
Amplitude
Connected account
Datadog
Connected account
TOOLS
Amplitude tools
The Amplitude tools this pairing puts in reach, each one running on your user's own connected account.
Actions top 8 of 11
-
Create Cohort
actionCreate a static cohort in Amplitude from an explicit list of user or Amplitude IDs viaPOST /api/3/cohorts/upload(the only REST-documented cohort creation endpoint; behavioral/dynamic cohorts cannot be created via REST). Use List Cohorts to find an existing cohort ID if updating. Example: call withcohortName="Power Users Q3",appId=849238,idType="BY_USER_ID",ids=["user@example.com"],owner="admin@example.com"-> returns{cohortId: "abc123"}. See the documentation.Writev0.0.2 -
Download Cohort File
actionDownload a completed cohort export's member list (step 3 of 3: request -> status -> download). Call Request Cohort Download then Get Cohort Download Status first, and only call this once status isJOB COMPLETED— calling it earlier will fail. Returns one record per member (amplitude_id/user_id, plus any requested user properties) — not the cohort's own metadata (name, size, definition); use List Cohorts for that. Example: call withrequestId="req_456"-> returns{requestId: "req_456", memberCount: 4200, returnedCount: 4200, truncated: false, members: [{amplitude_id: "123456789", user_id: "user@example.com"}, ...]}. See the documentation.Read-onlyv0.0.2 -
Get Cohort Download Status
actionCheck whether a cohort download job has finished (step 2 of 3: request -> status -> download). Call Request Cohort Download first to get arequestId. This action polls internally for up to ~35 seconds before returning, since Amplitude's export jobs commonly take 30-60+ seconds regardless of cohort size. If the returnedasync_statusis stillJOB INPROGRESS, call this action again with the same request ID (it may take a few calls) — do not call Download Cohort File untilasync_statusisJOB COMPLETED. Example: call withrequestId="req_456"-> returns{request_id: "req_456", cohort_id: "abc123", async_status: "JOB COMPLETED"}. See the documentation.Read-onlyv0.0.2 -
Get Event Segmentation
actionQuery event segmentation data (counts, uniques, and other metrics) for one or more events over a date range from the Amplitude Dashboard REST API. Use this to analyze how an event trends over time, optionally broken down by user properties. Example: call withevent={"event_type":"Purchase"},startDate="20240706",endDate="20240805",metric="uniques"-> returns{data: {xValues: ["2024-07-06", ...], series: [[42, 51, ...]]}}(one value per day per requested series). See the documentation.Read-onlyv0.0.2 -
Get Funnel Analysis
actionCompute conversion rates across an ordered (or unordered/sequential) set of funnel steps over a date range from the Amplitude Dashboard REST API. Provide one event definition per funnel step. Example: call withevents=["{\"event_type\":\"Sign Up\"}","{\"event_type\":\"Purchase\"}"],startDate="20240706",endDate="20240805"-> returns{data: [{stepByStep: [...], cumulative: [...], eventCount: 4200}, ...]}(one entry per step). See the documentation.Read-onlyv0.0.2 -
Get Retention Analysis
actionCompute retention (return rate) between a start event and a return event over a date range from the Amplitude Dashboard REST API. Example: call withstartEvent={"event_type":"_new"},returnEvent={"event_type":"_active"},startDate="20240706",endDate="20240805"-> returns{data: {series: [[...retention counts per interval...]], seriesLabels: [...]}}. See the documentation.Read-onlyv0.0.2 -
Get User Activity
actionFetch the event stream for a single user by their numeric Amplitude ID from the Amplitude Dashboard REST API. Each returned event defaults to just event_type and event_time; passfieldsto also get e.g.event_properties,user_properties,uuid,session_id,amplitude_id,device_id(event objects can carry many properties, so these stay opt-in). Amplitude's API caps each request at 1000 events with no cursor; this tool pagesoffsetforward automatically to collect up tolimitevents (set it above 1000, up to 5000, to fetch more than one page). Use Search Users first to resolve the Amplitude ID. Example: call withuser=12345678,limit=50-> returns{events: [{event_type: "Purchase", event_time: "2024-08-05 14:22:10"}, ...], userData: {...}, truncated: false}(truncated: truemeans more events likely exist beyond what was returned — raiselimit/offsetto fetch further). See the documentation.Read-onlyv0.0.2 -
List Cohorts
actionList all behavioral cohorts in the Amplitude project. Returns acohortsarray; each entry defaults to id, name, size, lastMod, appId. Passfieldsto get more, e.g.description,published,archived, or the largedefinition(the cohort's filter logic) /owners/viewers(email arrays). Use this to discover valid cohort IDs for Request Cohort Download (step 1 of 3, followed by Get Cohort Download Status and Download Cohort File, to fetch a cohort's member list). Example: call with no parameters -> returns{cohorts: [{id: "abc123", name: "Power Users Q3", size: 4200, lastMod: 1722873600000, appId: 849238}, ...]}. See the documentation.Read-onlyv0.0.2
Triggers
No Amplitude triggers are available yet.
TOOLS
Datadog tools
The Datadog tools this pairing puts in reach, each one running on your user's own connected account.
Actions top 8 of 13
-
Get Account Info
actionDetect the Datadog region for the connected account. Call this FIRST before any other Datadog tool if you do not already know the region. Returns the region domain (e.g.datadoghq.com) which must be passed as theregionparameter to Search Logs, Search Monitors, Get Metric Data, and all other Datadog tools. See the docsRead-onlyv1.0.2 -
Get Account Region
actionDetect the Datadog region for the connected account by validating the API key against each regional endpoint. Use this to discover the correct region before calling other Datadog actions. See the docsRead-onlyv0.0.1 -
Get Metric Data
actionQuery time-series metric data for analyzing trends and system performance. Query syntax:avg:system.cpu.user{*},sum:requests.count{env:prod} by {host},max:system.mem.used{service:web}. Bothfromandtoare POSIX timestamps in seconds. Use Search Metrics first to discover available metric names. Use Search Hosts to find valid host tags for scoping. After finding anomalies, use Search Logs to investigate related entries. See the docsRead-onlyv1.0.2 -
List Logs
actionGet a list of logs matching a filter query. Uses the GET endpoint with a higher rate limit (3600/hr vs 300/hr). See the docsRead-onlyv0.0.1 -
Post Metric Data
actionPost custom time-series metric data points to Datadog. Data appears in dashboards, monitors, and can be queried via Get Metric Data. Points: JSON object where keys are Unix timestamps (seconds) and values are numeric, e.g.{"1640995200": 1.0}. Use Search Metrics to verify a metric name exists, or post to a new name to create it. This is a WRITE operation that creates or appends data to a metric time series. See the docsWritev1.0.2 -
Search Dashboards
actionList and search Datadog dashboards. Returns dashboard IDs, titles, URLs, and metadata. Dashboard URL:https://app.{region}/dashboard/{id}where region comes from Get Account Info. Use alongside Search Services to find dashboards related to a specific service. See the docsRead-onlyv1.0.2 -
Search Events
actionSearch Datadog events: monitor state changes, deployment markers, error spikes, and infrastructure events. Filter bysources(e.g.nagios,docker),tags(e.g.env:prod,service:web), andpriority(normalorlow). Time range defaults to last 24h (POSIX timestamps in seconds). To investigate a monitor alert, use Search Monitors first, then search events for the relevant time range. Follow up with Search Logs for deeper investigation. See the docsRead-onlyv1.0.2 -
Search Hosts
actionSearch monitored infrastructure hosts. Filter by tag (env:production), name (host:web-01), or partial match. Sort bycpu,iowait,load,status, orapps. Host names from results can scope queries in Get Metric Data (e.g.avg:system.cpu.user{host:web-01}), filter logs in Search Logs (host:web-01), or filter metrics in Search Metrics. Max 1000 results. See the docsRead-onlyv1.0.2
MULTI-APP
Works with more apps
The combinations customers connect alongside these two — nothing here is limited to a pair.
REFERENCE
Toolkit details
- x-pd-app-slug
- amplitude,datadog
- Primary app
- Amplitude (amplitude)
- Second app
- Datadog (datadog)
- Authentication
- API key + API key
- Available actions
- 24
- Available triggers
- 1