Ad network MCP guide
Connect TikTok Ads to Claude or ChatGPT with Lemon AI’s unified MCP and API. Compare video performance and creative attributes with clear evidence.
TikTok advertising data can reach Claude through TikTok's own connector, or reach Claude and ChatGPT through Lemon AI's unified MCP. Choose the native route for TikTok-specific tools. Choose Lemon when the question spans your connected networks or needs creative performance together with the attributes assigned to each asset.
If you currently paste TikTok reports into chat, start by writing down the follow-up you cannot answer from that snapshot. It might be “show the same videos in another period” or “compare the opening hooks.” A connection is useful when it lets the assistant retrieve that evidence reliably.
Native TikTok connector or Lemon MCP
| Route | Suitable question | Boundary |
|---|---|---|
| Uploaded TikTok report | What changed in this saved report? | New periods and dimensions require another export |
| TikTok for Business connector | What can TikTok's own tools report or manage? | TikTok-specific access and operations |
| Lemon MCP and API | How do my creatives and their attributes perform across connected sources? | Read-only analysis using Lemon's available metrics |
The TikTok for Business listing in Claude's marketplace identifies TikTok as the connector's maker and includes reporting and management tools. Follow Add to Claude, sign in, and inspect the tools you enable. Begin with account discovery and a bounded reporting question before considering any mutation.
That listing documents a Claude connection. Do not assume its client-specific endpoint is a universal setup instruction for every AI app. The Lemon route below uses the same Lemon server URL for both Claude and ChatGPT.
Connect TikTok data through Lemon
In Lemon's Connections, add TikTok Ads and complete the provider connection flow for the account you intend to analyze. Select its data for sync, then confirm the campaign, asset types, and a complete date range in Creative Analytics. Open API & MCP to authorize the assistant.
These are two separate connections: TikTok supplies data to Lemon; the assistant queries Lemon's reporting tools. Signing the assistant in does not bypass TikTok account permissions or make an unsynced period available.
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.
Establish what the connected report can answer
Use Lemon AI to list my TikTok accounts and available date ranges. Let me choose the account and campaign. Compare video assets over two complete weeks using impressions, clicks, cost, and CTR. State unavailable metrics, missing labels, and truncation before drawing conclusions.
This delivery-first question avoids assuming that every native TikTok metric exists in the unified interface. A metric visible in Ads Manager is not automatically available through every connector. In particular, do not manufacture a post-install D7 result when account discovery does not advertise that window.
Lemon's tools can connect the returned asset to stored creative attributes using list_creative_attributes, attribute_insights, and get_creative. The attributes are the account's defined vocabulary, not whatever visual category the model happens to invent during the conversation.
A stronger click response is not a revenue result
Synthetic example, not customer data: two fictional videos each receive 10,000 impressions. A puzzle-reveal opening gets 180 clicks; a character-introduction opening gets 120. Their CTRs are 1.8% and 1.2% respectively.
| Fictional opening | Synthetic impressions | Synthetic clicks | Calculated CTR |
|---|---|---|---|
| Puzzle reveal | 10,000 | 180 | 1.8% |
| Character introduction | 10,000 | 120 | 1.2% |
The synthetic difference is 0.6 percentage points. The first video has a higher observed click rate in this example. Without attributed value or installs, the assistant cannot conclude that it has better ROAS or CPI. Two videos also cannot establish that the hook category itself caused the difference.
This is where an analysis should become more specific rather than more confident: check delivery scope, examine more independent creatives, and determine which downstream outcomes are actually available. The creative-ranking reliability method offers a reader-run check for a ranking before acting on it.
Ask for a hook investigation and a cross-network follow-up
For the same TikTok account and period, inspect the available hook labels with attribute_insights. Use CTR, retain evidence qualification, and show coverage and examples. Separate what the data supports from a proposed explanation.
Compare this hook hypothesis with my selected Meta account using the same country, video format, and dates where available. Keep network results separate. Explain differences in coverage and metric definitions before proposing one new video test.
This is the practical reason to use a unified MCP: the assistant can discover and query another connected source without asking you to export another table. It still needs to establish that the comparison is meaningful. Similar labels and column names do not guarantee comparable audiences.
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.
When the connection works but the answer is incomplete
Check whether the assistant selected the right account and used a synced date range. For recent data, a partial day can look like declining delivery. If the requested video metric is not exposed, ask for the available columns instead of accepting a guessed substitute.
For native TikTok tools, distinguish reporting permissions from management permissions. For Lemon, the MCP analysis tools cannot edit the campaign. Read-only access still shares reporting data with the AI app you authorize.
Connect your ad accounts to build the recurring workflow. The MCP connection guide covers the broader architecture, and Reports explains the API and reporting views.