The TikTok for Business MCP Server is TikTok's own connector between AI assistants and TikTok Ads, and it is quick to put in Claude: it is listed in Claude's connector directory, and connecting is a TikTok for Business sign-in, with no developer app or API key. It also writes. Its tools create campaigns, move budgets and bids, pause ad groups, and now run creative tests. Below: setup in Claude and Claude Code, the tools that matter for analysis, the limits, a five-step weekly workflow, and what changes when the analysis is done before the chat starts.
Facts about TikTok's server were checked against TikTok's developer documentation, help center, and newsroom, and Anthropic's connector directory, on November 3, 2026.
What is the TikTok Ads MCP server?
It is a Model Context Protocol server that TikTok hosts, turning about 400 TikTok API for Business endpoints into tools an AI assistant can call: campaign management for Upgraded Smart+ and GMV Max, reporting, catalogs, creative assets, audiences, and Business Center. TikTok's changelog dates the launch to July 2026, its first agent tool, a diagnosis agent, to August, and a creative testing agent to October. At Advertising Week New York in October, TikTok announced it as a ready-made connector in Claude and seven other AI platforms and registries.
It is not the only way to get TikTok into Claude. There are four common routes:
| TikTok's MCP server | Open-source local server (AdsMCP) | Hosted connector (Pipeboard) | Peachblue | |
|---|---|---|---|---|
| Where it runs | TikTok's servers | Your machine | Pipeboard's servers | Peachblue's servers |
| What it covers | TikTok Ads, about 400 tools | TikTok Ads, 20 read-only tools | TikTok, Meta, Google Ads, Snap, Reddit | TikTok, Meta, Google Ads, and Amazon DSP |
| What comes back | API results, plus plain-language diagnosis and test reports | API results | API results | Rankings, video tags, and creative analysis, computed |
| Can change your account | Yes: campaigns, budgets, bids, audiences, creative assets | No | Yes, with permissions; read-only tokens from Pro | No, read-only |
| What you need first | A TikTok for Business account | Your own TikTok developer app, Python, and uv | A Pipeboard account | A Peachblue plan and a TikTok sign-in |
| Cost | No fee listed | Free, open source (MIT) | Free plan; Pro $29.90/mo | TikTok from $79/mo (Starter); Claude access from $199/mo (Pro) |
Pipeboard's prices were verified October 24, 2026; AdsMCP's facts come from its README, read November 3, 2026.
How do you connect TikTok's MCP server to Claude?
Add it from Claude's connector directory and sign in with your TikTok for Business account. There is no developer app, API key, or config file, and TikTok says the server is publicly available to all users, with no permissions to apply for.
- Find it. In Claude, open Customize, then Connectors, and search for TikTok. The listing is "TikTok for Business", published by TikTok. If you start in Claude Desktop, Claude may send you to Claude on the web to finish.
- Authorize it. Click connect, sign in to TikTok Ads Manager if asked, review the permissions, and click Authorize.
- Test it. TikTok suggests a first prompt: "Show me the performance report for all active campaigns this week."
The directory version leaves out 26 tools, mostly billing and payments (transactions, invoices, balances, transfers, payment portfolios), plus TikTok's diagnosis agent (TikTok's list). If you need those, add the server by URL instead: in Customize, Connectors, add a custom connector with a name and one of two URLs.
- Full Disclosure,
https://business-api.tiktok.com/open_mcp/tt-ads-mcp-flat: all of the roughly 400 tools load at connection. TikTok's help center recommends it for Claude. - Progressive Disclosure,
https://business-api.tiktok.com/open_mcp/tt-ads-mcp-layer: about 40 core tools load first, and the rest are found on demand. Use it to save context.
In Claude Code it is one command, then /mcp to sign in (Claude Code docs):
claude mcp add --transport http tiktok-ads https://business-api.tiktok.com/open_mcp/tt-ads-mcp-layer
On plans, the two companies differ. TikTok's Claude guide asks for a Pro or Max plan, while Anthropic's help center lists custom connectors, the URL route, on every plan: one on Free, and on Team and Enterprise an owner adds it first.
Three cautions. First, the sign-in lasts 30 days; after that you authorize again with the same TikTok for Business account. Second, this server can spend money: keep Claude's approval prompt on for its tools rather than allowing them always, and agree who signs off on budget changes before anyone types "scale it". Third, TikTok allows 6 queries a second per tool per user, says tools vary by region and account setup, and asks for no more than one edit every five seconds to a single Smart+ ad (limits).
Peachblue's connector goes in as a custom connector with no developer setup: copy the URL from Settings, MCP tab, add it, and approve the sign-in (setup docs). Connecting TikTok to Peachblue is a TikTok for Business sign-in where you pick the advertiser accounts to sync, and nothing in them is ever changed (TikTok setup).
Which TikTok MCP tools matter for analysis?
A handful of the 400 do the analytical work: the standard report, Video Insights, in-second retention, ad benchmarks, and TikTok's two agents. Each returns TikTok's numbers about a video. Reading the footage is left to Claude, and Claude does not read video files.
report_integrated_get is the standard report, by campaign, ad group, or ad. One trap: TikTok counts conversions on the date of the impression, while separate "real-time" metrics count them on the date of the conversion (metric definitions), so ask for one and say which. Peachblue syncs daily and reads the previous seven days again on every run, so late conversions land without a new query.
creative_report_get runs a Video Insights report across many videos and ad accounts, Spark Ads included: spend and results beside paid and organic views, 2-second and 6-second views, 25, 50, 75, and 100 percent completions, average play time, and likes, comments, and shares. That is everything a TikTok hook rate needs, 2-second views over impressions, as our hook rate guide explains. Peachblue adds the part the numbers cannot: what is in the video.
report_video_performance_get gives in-second data for a video's first 60 seconds: how many viewers stay, drop off, click, and convert, second by second, about an hour behind (reference). It is the best tool in the set for finding where a video loses people. It shows when viewers leave, not what was on screen at that moment; Peachblue analyzes each video frame-wise across its timeline, so the hook, the pacing, and the scenes are part of the read.
report_ad_benchmark_get compares an ad with benchmarks by location, ad category, conversion event, and placement over 7 to 60 days. It needs 1,000 impressions per ad in the window and runs 30 to 48 hours behind. Peachblue scores each creative from 0 to 100 against the rest of your account and within TikTok, so the benchmark is your own account.
The two agents. tiktok_ads_diagnosis_agent is TikTok's own troubleshooter for accounts, campaigns, and ad groups outside GMV Max, and it answers in plain language; it comes only with the URL setup. tiktok_ads_creative_testing_agent drafts a creative test inside a live ad group, waits for "Start the test", and ends with per-creative results and a recommended action (TikTok's guide). Its rules are strict: videos only, an Upgraded Smart+ ad group on Maximum Delivery with a daily budget and spend in the last 24 hours, 1 to 50 creatives, at least three days, and a test budget that is not a hard cap.
Creative. TikTok's documentation describes no tool that tags what is in a video (hook style, format, tone) or matches a TikTok video to the same creative on Meta, Google Ads, or Amazon DSP. Peachblue tags every TikTok creative on 31 dimensions, writes an analysis with copy ideas, and recognizes the same creative across ads and platforms, even after re-crops and re-compression.
How do you analyze TikTok ads in Claude, step by step?
Five weekly steps: pin the account, pull performance, find the hook and the drop-off, judge the creative, then explain the change and test the next round. Each works through TikTok's server; a layer that has already done the math changes how much lands on the chat.
1. Pin the account
Use advertiser [ID] for everything in this chat, and tell me before you switch.
One sign-in can reach many advertisers, especially through a Business Center, so name the one you mean.
In Peachblue the scope is the advertiser accounts you synced, the same ones your team sees in the app.
2. Pull performance for the period
Pull an ad-level report for the last 14 days: spend, impressions, clicks, conversions, and cost per conversion, counted by impression date. Rank by spend and flag any ad with spend and no conversions.
Check one ad against TikTok Ads Manager before trusting the rest.
In Peachblue this is get_campaigns and get_performance_summary, from data synced daily, with get_risk_summary flagging zero-result spend.
3. Find the hook and the drop-off
Run a Video Insights report for our 10 videos with the most spend: impressions, 2-second and 6-second views, and 25, 50, 75, and 100 percent completions. Give each a hook rate (2-second views divided by impressions). Then pull in-second retention for the top three and name the second where most viewers leave.
A TikTok hook rate built on 2-second views is not comparable to a Meta one built on 3-second plays, so keep the two apart.
In Peachblue every video was analyzed frame-wise across its timeline when it synced, and get_creative_detail hands Claude its tags, placements, and a written read of what the opening does.
4. Judge the creative, not the ad
Combine ads that use the same video, then rank the videos by cost per conversion with spend beside each. What do the top five have in common?
Claude can combine rows by TikTok's video IDs, but it answers "what do they have in common" from names, captions, and numbers, because it does not watch the videos, and it never sees your Meta cut of the same creative.
In Peachblue this happened at sync: every creative is fingerprinted, grouped across ads, and matched to the same creative on Meta, Google Ads, and Amazon DSP. get_creatives ranks them by composite score, and get_patterns names what your winners share.
5. Explain the change and test the next round
Compare the last 7 days with the 7 before. Which ads and videos moved cost per conversion most, and by how much? Write a five-line update using only numbers from this chat.
If you run Upgraded Smart+, ask TikTok's creative testing agent to draft a three-day test of your new videos in the ad group with the steadiest spend, and read the plan's budget before you reply "Start the test".
In Peachblue this is compare_periods, which names the creatives that moved the result, and the Next Creative Brief, which turns your winners' patterns into concepts for the next videos.
What makes Peachblue better than TikTok's MCP server for creative work?
The analysis is finished before the conversation starts, the video itself is part of it, and TikTok sits beside Meta, Google Ads, and Amazon DSP in one connection. TikTok's server gives Claude the controls and the numbers; Peachblue gives it answers about your creative:
| You ask Claude | With TikTok's server | With Peachblue |
|---|---|---|
| "Which TikTok video is winning, and why?" | Video Insights rows; no tool reads the footage | Ranked by one composite score, tagged on 31 dimensions, with a written read |
| "Why do viewers drop off?" | The second they leave, for the first 60 seconds | The hook and pacing, read from the video and tagged |
| "Is our best TikTok also our best Meta ad?" | TikTok only | The same creative matched across platforms |
| "What changed this week?" | Two reports and the math in chat | compare_periods names the creatives behind the change |
| "What should we make next?" | A draft from whatever is in the chat | A brief built on your winners: concepts, proven hooks, a prompt block |
Performance. Claude reasons over results instead of assembling reports. get_daily_spend gives day-by-day totals for any window, get_risk_summary flags spend concentration, fatigue candidates, and zero-result spend, and questions over MCP use no credits.
Creative. Every TikTok video is analyzed frame-wise across its timeline, tagged on 31 dimensions, given a written analysis and copy ideas, and scored against your whole account and within TikTok. Because the same creative is matched across platforms, a TikTok winner and a Meta winner meet in one chat. It is the only creative analytics MCP that serves your own cross-platform performance data, Amazon DSP included.
The loop. The Next Creative Brief turns your winners' patterns into concepts and proven hooks, with a prompt block for an image or video generator. TikTok's testing agent can then give the new videos a controlled budget inside a live ad group, and once they have run, Peachblue scores them on the same scale as everything before.
TikTok's server keeps what is its alone: it runs the account. Building Smart+ and GMV Max campaigns, moving budgets and bids, uploading creative, managing catalogs and audiences, second-by-second retention, and live creative tests all happen there. Connect both, one to act and one to know what to make next.
Which route should you take?
- You want Claude to run TikTok campaigns, from building them to budgets and tests: TikTok's server, from the directory. Keep approvals on and agree who signs off on spend.
- You want read-only TikTok data on your own machine: AdsMCP's open-source server, with your own TikTok developer app.
- You run TikTok beside Meta, Snap, or Reddit and want one hosted connector that can also make changes: Pipeboard.
- You want to know which TikTok creatives win, why, and what to make next, beside Meta, Google Ads, and Amazon DSP: Peachblue. TikTok analysis starts on Starter at $79 a month, Claude access on Pro at $199, and the same tools run in the product as Agent Peach.
One boundary, stated plainly: Peachblue reads. It does not create, edit, pause, or test anything in TikTok Ads Manager. Every server is compared in the best MCP servers for ad data; the Meta, Amazon, and Google versions of this guide are Claude for Meta ads creative analysis, Amazon Ads MCP in Claude, and Google Ads MCP server in Claude; the four-platform workflow is Claude for media buyers.