GET ON THE MCP WAITLIST
Run your Amanda AI ads from Claude
Amanda AI is launching an MCP that connects our platform to Claude. Create and run your ads straight from Claude Desktop, with Amanda AI's optimization and execution layer between Claude and your ad account.
How it fits together
Your ad accounts connect to Amanda AI, Amanda AI connects to Claude, and you run your ads through Claude. We do the calculations on our backend and send Claude only what’s relevant to what you’re doing.
Google Ads
Meta Ads
Bing Ads
Amanda AI
Spend limits – Optimization – Site & feed crawls
Claude Desktop
Three layers between Claude & your budget
Claude never touches your ad accounts directly. Here’s what sits in between.
Rules
Your spend limits are set in Amanda AI, not in the conversation. Every request from Claude gets checked against your daily thresholds and keyword limits before anything happens.
Data
Amanda AI imports your full account history and crawls your landing pages, product feeds, and competitors. Years of logs and messy site structure get compressed into something Claude can actually reason about, instead of drowning in it.
Optimization
Keyword mining, continuous target evaluation, and creative rotation on Meta. Conversion lag is handled per channel on Google, so ad groups aren’t punished for conversions that arrive late.
Why run your Amanda AI ads with Claude?
Combining the world’s best language interface with Amanda AI, a high-speed mathematical execution engine, gives you correct calculations, not guesses.
Chat and build
Tell Claude what you’re launching, which platforms it should run on, and what target to hold. Amanda AI sets it up and takes over from there. Checking performance works the same way: ask, and the numbers come back calculated rather than estimated.
Three platforms, one chat
Connect Google, Meta, or Bing. All three run in one conversation. Amanda AI moves budget between them based on what’s actually working, so you don’t have to reallocate manually every time performance shifts.
Claude stays out of the math
LLMs reason with language. They don’t count. Budget decisions, bidding, and pacing stay with Amanda AI, which returns the same answer to the same question every time.
Your data arrives compressed
Dumping account history and product feeds straight into Claude burns context and slows everything down. Amanda AI does the processing first: about 2,500 tokens instead of two million.
Benchmark
Two million tokens in. 2,500 out.
Amanda AI does the work before Claude gets involved. We pull your account data, run the calculations on our backend, and send Claude the results. Not the data behind them.
One brand, 400 active items tracked continuously across Google, Meta, and Bing: product catalog rows, performance history, conversion tracking, and search query logs.
The cost-saving is the obvious part and the part that matters least. The real problem with two million tokens is that models get worse as context gets longer. Details go missing. A short, structured payload doesn’t just cost less, it gets you a better answer.
*The structured payload Amanda AI passes to Claude on each run. Contains the resolved figures, not the underlying account data.
**Difference between the raw payload and the structured one, as a percentage.
*** ?
64,400,000
Tokens saved per month ***
2,150,000
Tokens if you push your raw account data into Claude yourself *
2,500
Tokens Amanda AI sends instead, per run *
99.88 %
Of the raw data never reaches the chat window **
