Ad network MCP guide
Connect AppLovin to Claude or ChatGPT with Lemon AI’s unified MCP and API. Keep independent asset estimates separate from source totals.
Connect AppLovin to Lemon AI, then add Lemon's MCP server to Claude or ChatGPT to query creative performance without repeatedly exporting reports. The same connection exposes your other connected ad networks and the creative attributes stored in Lemon. For AppLovin, the important distinction is between network-reported delivery and Lemon's estimated asset outcomes.
This guide concerns advertiser-side AppLovin and Axon analysis. AppLovin MAX monetization reporting answers a different question. A connector that reads MAX revenue is not automatically a connector for evaluating the ads you buy.
Choose the reporting route for the question
| Route | What it gives the assistant | Work that remains |
|---|---|---|
| Paste an Axon export | A fixed report with the columns you selected | Repeat exports and preserve the report's grain and definitions |
| Build against AppLovin's reporting APIs | Direct access to the reports you implement | Authentication, report selection, storage, and an API tool or MCP wrapper |
| Use Lemon's MCP and API | Synced creative reporting, estimated asset outcomes, and assigned attributes | Choose the account and window; inspect estimation and aggregation limits |
AppLovin's Asset Reporting API is a primary source for asset delivery. If you build a direct agent tool, keep asset and campaign or creative-set reports separate rather than treating every endpoint as the same table. The AppLovin creative-reporting guide covers the measurement boundary in more detail.
For a direct API workflow, request a small date range at one explicit report grain, retain the source fields, and let your own tool handle authentication. An API URL by itself is not an MCP server URL. Never paste an AppLovin report key into an AI conversation.
Connect the AppLovin account in Lemon
- Open your Axon account settings and identify the Account ID.
- Obtain the Reporting API Key from the account's API keys page.
- In Lemon, open Connections, choose AppLovin, and enter the account ID and reporting key in the connection form. Supply the promoted website domain if it applies to your setup.
- Select the accounts to sync and confirm that the intended app, creative assets, and dates appear in Creative Analytics.
- Open API & MCP and authorize your assistant separately.
An assistant can then ask which creative files performed differently and inspect the labels attached to those files. Lemon's attribute analysis connects that reporting question to the hook, visual treatment, or format you may want to test next.
Add Lemon to Claude or ChatGPT
Use the same remote MCP server for all the networks you connect in Lemon:
https://api.lemon-ai.com/mcp
In Claude, open Customize → Connectors → + Add → Add custom connector. Enter a name and the URL, continue, and sign in to Lemon. If Claude asks how to register the OAuth client, choose Register automatically. Enable the connector in the conversation. Organization-managed accounts may need an owner to add it first. See the Claude setup instructions.
In ChatGPT, open Plugins, choose Add custom MCP server, and enter the URL. Select OAuth authentication, authorize Lemon, and install the resulting personal plugin. Mention it in a Work chat to use its tools. Availability depends on your workspace settings; follow the current ChatGPT MCP instructions if the interface differs.
Start by asking the assistant to call list_ad_accounts and let you select the intended account. Then use list_filter_values to discover dates and filters. Your ad-network sign-in connects data to Lemon; this second authorization lets the AI app read it. Do not paste network credentials into chat.
Ask for a mature asset comparison
Use Lemon AI to discover my AppLovin accounts and available outcome windows. For the account and app I choose, find a complete two-week period with mature D7 data. Compare video assets using cost and the available D7 outcomes. Identify every estimated value, preserve non-additivity warnings, and show the selected dates. Do not total the asset rows into campaign revenue.
The agent should call list_ad_accounts before selecting roas:d7, then inspect cohort_metadata in the performance result. If partial is true, the window is still maturing. If the account does not expose the requested outcome, the assistant should explain the gap.
MCP makes a report easier to retrieve; it does not change the source or turn an estimate into a network-reported observation.
Why summing assets can produce the wrong answer
AppLovin asset outcomes in Lemon are estimated independently. They are marked non-additive, so a collection of asset rows must not be treated as a partition of campaign revenue.
Synthetic illustration, not customer data or a reconciliation result: imagine a creative set with $100 of source-reported value. Two participating assets have illustrative estimated values of $70 and $60.
| Fictional row | Synthetic value | Interpretation |
|---|---|---|
| Creative set | $100 | Source-report total in this illustration |
| Gameplay video | $70 | Independent illustrative asset estimate |
| Playable | $60 | Independent illustrative asset estimate |
The asset sum is $130. That is not evidence of an extra $30 earned by the campaign. It is a warning that the rows do not define mutually exclusive parts. Use the matching source-report grain for the campaign-total question and asset-grain evidence for the asset-comparison question.
The same care applies to delivery totals and ratios. Do not mix a numerator from one report with a denominator from another simply because both columns have familiar names. Read the asset-level creative ROAS method before turning asset rankings into budget decisions.
Use creative attributes to choose a test
In the same AppLovin account and mature period, list the creative attributes. Investigate CPI by the available hook label with attribute_insights. Show evidence coverage and examples, and state which values are estimated. Keep the default evidence qualification.
Take the apparent hook advantage and check whether it persists within the same campaign and country. Produce one creative-test brief: what to preserve, what to change, and which result would reject the hypothesis. Do not change the campaign.
Ask for examples through get_creative when the finding needs inspection. A label that groups several different execution styles may need refinement before it can guide a production brief.
Use the same reporting layer through the API
For a script or another AI app, create a Lemon API key in API & MCP. Retrieve your account IDs with GET /v1/ad-accounts, then make a scoped read. This example uses a placeholder account and dates; substitute values returned for your account.
curl --get 'https://api.lemon-ai.com/v1/creatives/performance' \
--header "Authorization: Bearer $LEMON_API_KEY" \
--data-urlencode 'ad_network_account_id=YOUR_ACCOUNT_ID' \
--data-urlencode 'date_from=2026-10-01' \
--data-urlencode 'date_to=2026-10-07' \
--data-urlencode 'metric=impressions' \
--data-urlencode 'metric=clicks' \
--data-urlencode 'metric=cost' \
--data-urlencode 'sort_by=cost' \
--data-urlencode 'sort_order=descending' \
--data-urlencode 'limit=25'
MCP and REST expose Lemon's reporting definitions. Inspect missing fields, result limits, and any outcome metadata before interpreting a result. A limited top-creatives response is not an account total. Use the API documentation for supported fields and pagination.
Troubleshoot the report before changing the creative
If the account is missing, verify the Axon ID and reporting key in Lemon. If recent rows are absent, inspect sync status and available dates. If D7 looks incomplete, compare the report watermark with the maturity date. If a sum looks too large, check the reporting grain and non-additivity flag before assuming ingestion duplicated revenue.
For a consistent starting point, connect your ad accounts. Use Reports and API for the interfaces and the worked MCP analysis for the sequence from evidence to a test.