Muze AI
Muze AI is an autonomous performance-marketing operator that researches, builds, launches and optimizes paid campaigns across Meta, Google, Amazon and Shopify, either approving every change or running on Autopilot, from a dashboard or any MCP client.
What is Muze AI?
Muze AI is an autonomous performance-marketing operator built to run paid advertising end to end rather than assist with one slice of it. The product, called Muze CMO in its legal documents, joins research, campaign creation, launch and continuous optimization so the work does not break across separate tools and teams. The homepage frames this as four stages: discover, create, launch, optimize. It reaches three advertising platforms and one commerce source. On Meta it handles campaign and breakdown analysis, creative intelligence, catalog ads, audience research, campaign creation, budget optimization, image and video creative, and an Autopilot. On Google it spans Search, Performance Max, Shopping, Display, Demand Gen, Video and App campaigns, keyword intelligence and conversion health. Amazon adds profiles, reports, search terms, targeting and bids, while Shopify supplies read-only product, order and revenue data used for blended reporting. TikTok and Bing appear on the full plan. The tool exists on two surfaces. One is a web dashboard. The other is a hosted Model Context Protocol server exposing 149 tools — eight cross-platform, 37 for Meta, 41 for Google, 48 for Amazon and 15 for Shopify — to an AI client the customer already uses, such as ChatGPT, Claude, Cursor, Codex or Slack. That server is a passthrough to the underlying advertising APIs and does not run Muze's own analysis; customers bring their own assistant and are never billed for model tokens. Control is the design point. Any tool that changes something returns a preview first and executes only on a second, explicit call carrying a confirmation. Anything created is staged paused and cannot spend until a human activates it, and permissions are scoped account by account with a full history of recommendations and actions. The company is young: the domain was registered in August 2025 and first archived in December 2025, and the privacy policy names AgentProd Inc., San Francisco, as controller of record. Headline figures such as more than 100 million USD in managed ad spend and a 3x average ROAS are vendor claims, not audited results.
What it does
- Analyze connected Meta, Google, Amazon and Shopify accounts and rank the issues by wasted spend
- Build complete campaigns with objective, audience, budget, copy and creative, staged paused for review
- Generate image, video, UGC and B-roll ad variations from a brief or a product URL
- Publish approved campaigns directly to Meta and Google
- Pause losing ads and scale winners automatically within rules the customer defines
- Detect creative fatigue three to five days before performance drops
- Answer paid-media questions from ChatGPT, Claude, Cursor, Codex or Slack through 149 MCP tools
When to use Muze AI / When not to
A quick filter to help you decide if Muze AI is the right fit.
When to use Muze AI
- DTC and e-commerce brands running Meta, Google or Amazon ads without an in-house media buying team
- Performance marketing agencies handling several client ad accounts from a single place
- Founders and small growth teams whose advertising budget starts around 200 USD a month
- Shopify merchants who need catalog sync, product-level ROAS and feed errors fixed
- Technical teams who would rather drive paid media from ChatGPT, Claude, Cursor or Codex than learn another dashboard
When not to use Muze AI
- Advertisers with no existing Meta, Google, Amazon or Shopify account to connect
- Anyone expecting the vendor to fund, front or bill their media spend, which stays on their own platform accounts
- Buyers looking for a guaranteed advertising outcome, which the terms explicitly refuse to promise
- Organizations that require EU data residency or a security certification such as SOC 2 before onboarding a vendor
- Teams wanting write actions at no cost, since the free MCP tier is strictly read-only
How to use Muze AI
A typical end-to-end flow, from setup to results.
- Create a Muze account and pick a plan, starting from the free read-only MCP tier if you only want to evaluate
- Authorize the platforms you want covered, Meta, Google, Amazon or Shopify, through each provider's OAuth flow
- Let Muze run its opening account audit and review the issues it ranks by wasted spend
- On Muze Lite, answer the five-minute questionnaire about your business, offer and goals so the AI can frame a strategy
- To work from an AI client instead, add the Muze MCP endpoint as a remote connector in Claude.ai or ChatGPT and complete the OAuth prompt
- In desktop or developer clients such as Cursor, Claude Code or Codex, create an API key on the API access page and send it in an X-API-Key header
- Ask a first bounded question, for example listing your connected accounts and summarizing the last seven days
- Read the preview Muze returns before confirming anything, since no change is written without an explicit confirmation
- Confirm the actions you want, then activate the campaigns Muze created in a paused state to start spending
- Set the budgets, permissions and rules if you want to hand routine optimization over to Autopilot
Pros & Cons
Pros
- Covers three advertising platforms plus Shopify commerce data in a single operator, which is unusual in this category
- Serious execution guardrails: a preview before every write, an explicit confirmation, and new campaigns staged paused
- Flat subscription pricing with no cut of advertising spend, except on the Enterprise tier
- A permanent free read-only tier allows the whole surface to be evaluated without a credit card
- Customers bring their own AI assistant and are never charged for model tokens
- An unambiguous commitment never to train any model on customer campaign, store or creative data
- A named list of ten subprocessors is published, alongside encryption, key hashing and per-account isolation details
Cons
- A very young vendor: the domain dates from August 2025 and the site was first archived in December 2025
- Corporate transparency is thin, with no street address, no about or contact page, no social profiles and one email address for legal, privacy and support
- Three names circulate for one service: the Muze AI brand, the Muze CMO product and AgentProd Inc. as controller of record
- No Article 27 EU representative is named, and the privacy policy still contains an unresolved note asking counsel to confirm its transfer mechanisms
- No security certification is claimed, and hosting is United States only with no EU data residency option
- Subscriptions and consumed tool units are non-refundable, including for the time left after a cancellation
- Evidence is limited to a single documented case study, with homepage performance tiles that are illustrative rather than audited
Pricing & Plans
A permanent free plan is available: the Muze MCP Free tier costs 0 USD per month and provides 25 read-only tool calls across Meta, Google, Amazon and Shopify, with no credit card required. The lowest paid entry point is Muze MCP Starter at 29 USD per month. Dashboard plans are billed separately, beginning with Muze Lite at 99 USD per month, while the AI Performance Marketer plan is 399 USD per month and Enterprise starts at 999 USD per month plus 1% of monthly advertising spend. Every paid plan opens with a free trial and is covered by a 14-day guarantee, is billed monthly in advance through Stripe, and may be cancelled at any time.
- 25 tool calls
- read-only across Meta
- Amazon and Shopify
- works with any MCP client
- 400 tool units and confirm-gated writes on Meta
- Google and Amazon
- Shopify remaining read-only
- 2
- 500 tool units
- the highest rate limits
- priority support and multi-account work
- AI video and image creative
- Meta and Google connections
- one-click campaign launch
- instant account audit and competitor ad cloning
- everything in Lite plus 400 credits
- 24/7 autonomous optimization
- Creative Cortex
- fatigue detection and Performance Genome
- unlimited credits
- white-label platform
- multi-account dashboard
- dedicated account manager
- two-hour priority support and SLAs
Data, GDPR & hosting
A consolidated view of how Muze AI handles your data.
GDPR overview
GDPR is addressed concretely rather than in passing. The privacy policy, effective 20 July 2026, carries a dedicated section on legal bases for processing that names four of them: performance of a contract, legitimate interests, legal obligation and consent. Rights of access, correction, deletion, export, objection, restriction and withdrawal of consent are set out for people in the EEA and the United Kingdom, exercised by email, with CCPA and CPRA rights covered alongside. Transfers to the United States rely on Standard Contractual Clauses, the UK Addendum where applicable, or an adequacy decision, and a Data Processing Agreement is offered on request. Two gaps stand out: no Article 27 EU representative and no data protection officer are named, and the transfer section still carries an unresolved bracketed note asking counsel to confirm which mechanism is actually in place with each subprocessor.
Who owns the data?
Under the terms, the customer keeps every right to their own material: account information, advertising data, store data, and the campaigns and creatives produced through the product. Muze receives only the limited licence needed to host, process and display that material in order to deliver the service, plus the right to pass it to the subprocessors named in the privacy policy for that same purpose. Creatives generated for a customer are theirs to run in their advertising. Legally, Muze acts as data controller for the account details supplied directly and as service provider or processor for the ad and store data it fetches on the customer's behalf; the controller of record is AgentProd Inc.
Reuse rights
Yes. Campaigns and creatives generated inside Muze belong to the customer and can be used in their advertising without asking further permission; Muze retains rights to the software, dashboard, MCP server and brand, not to the output. Muze restricts its own reuse tightly: it does not sell, rent or trade personal or advertising data, does not train its own or any third party's models on customer data, and does not blend one account's data into another customer's analysis or benchmarks. Data obtained from Meta, Google, Amazon and Shopify is used solely to provide the features the customer switches on, never for Muze's own advertising and never to build audience profiles outside that account.
Data retention & training
Hosting summary
Muze states that it operates in the United States, and its subprocessors process data in the United States and, in some cases, other countries that are not named. There is no EU data residency option. Ten subprocessors are listed: Supabase for the application database and authentication, Amazon Web Services S3 for storing generated creatives, Redis for caching and usage metering, Stripe for payments, OpenRouter, Google's Gemini API and OpenAI for language model analysis and generation, FAL for image and video creation, and PostHog and Amplitude for product analytics. An AI assistant connected over MCP is treated as the customer's own processor rather than a Muze subprocessor. Where transfers need additional safeguards, Muze relies on Standard Contractual Clauses with the UK Addendum where applicable, or an adequacy decision, although the policy still flags that section for legal confirmation. Security measures described include HTTPS and TLS in transit, MCP API keys stored only as SHA-256 hashes, Shopify tokens encrypted at rest, scoped OAuth with minimum permissions, and per-account data isolation.
Things to keep in mind
Risks and trade-offs to weigh before adopting Muze AI.
- The tool makes real changes to accounts that spend real money, and the confirmation step is the only barrier; clicking through previews out of habit quietly removes it
- Actions taken through a connected AI assistant are contractually your actions, including when the assistant misreads what you asked for
- Campaigns arrive paused, but activating one commits budget immediately, so a routine approval can become unplanned spend
- Handing optimization to Autopilot can erode a team's own feel for what is working, leaving nobody able to judge the machine's calls
- Figures shown in the product come from the platform APIs and may be delayed, restated or different from the native interfaces, so financial decisions should be checked at the source
- No advertising performance is guaranteed and liability is capped at the last twelve months of fees, so the cost of a bad run stays largely with the customer
- Data is processed in the United States with no EU residency option, and API keys are displayed once and cannot be recovered if lost
Setup & Integrations
Technical difficulty
Low for most users, moderate on one path. Through the dashboard nothing is technical: create an account, authorize Meta, Google, Amazon or Shopify over OAuth, and Muze audits the account immediately; Muze Lite adds a five-minute questionnaire and a two-click Meta connection. Hosted AI clients such as Claude.ai and ChatGPT need only the MCP connector URL and an OAuth approval. The step requiring some comfort with configuration is desktop and developer clients, where you create an API key and paste a small JSON block carrying an X-API-Key header. Existing ad accounts and the rights to authorize them are prerequisites.
Deployment
Integrations
Supported languages
Behind Muze AI
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Muze AI.
Frequently asked questions
Which advertising platforms does Muze AI work with?
Can Muze change or spend anything without my approval?
Is there a free plan?
How much does the cheapest paid plan cost?
Does Muze train AI models on my advertising data?
Can Muze see my conversations with my AI assistant?
Who pays for the advertising itself?
Where is my data hosted and how long is it kept?
Is a Data Processing Agreement available?
Should you pick Muze AI?
Muze AI is one of the more complete attempts to hand paid media over to software rather than to a person. Coverage is genuinely broad, spanning Meta, Google and Amazon advertising plus Shopify commerce data, and the two surfaces suit different buyers: a dashboard for teams who want an interface, and a 149-tool MCP server for those who would rather work inside ChatGPT, Claude or Cursor. The commercial model is legible, a flat subscription with no percentage of ad spend outside the Enterprise tier, and the free read-only tier lets the whole surface be assessed at no cost. What makes it credible is the plumbing around execution. Every write returns a preview, nothing changes without an explicit confirmation, and anything created arrives paused so it cannot spend before a human activates it. Combined with per-account permissions and a full action history, that is a more careful design than most tools acting directly on live budgets. The reservations concern the vendor rather than the product. This is a very young company: a domain from August 2025, a first web archive in December 2025, one documented customer story and illustrative homepage performance tiles. Corporate disclosure is thin, with no street address, no about or contact page, a single email for legal, privacy and support, and three names in circulation for one service. GDPR is addressed seriously, but no Article 27 representative is named, no security certification is claimed, hosting is United States only, and the published policy still carries a bracketed note asking counsel to confirm its transfer mechanisms. For a DTC brand or agency already spending on Meta and Google, it is worth a trial on the free tier. For a regulated or European buyer with vendor-assurance requirements, the paperwork is not yet there.
- Choosing a selection results in a full page refresh.
- Opens in a new window.