MarginTide: price intelligence that shows its evidence

Rivals change prices every day, and checking their storefronts by hand never keeps up. The same goes for supplier quotes, competitor plans and ad spend: the change is invisible until the invoice lands or the budget is already spent.

MarginTide is competitive pricing intelligence. It watches every price that sets your margin: what rivals charge, what suppliers charge you, how competitors package their plans, and what each sale costs to win. Then it recommends a move you approve, with the evidence attached. It is the first agent from our studio to become a product, and it grew out of the Price Checker Agent we built for client work.

Four ways margin leaks.

MarginTide is one product with four workflows. You start with the one where margin hurts most today, and add the others later on the same plan.

• Price Checker. Watches the retailers you compete with and turns every meaningful price move into a graded hold, match or move recommendation you approve.
• Procurement Intelligence. Benchmarks what your suppliers charge against real market prices, so savings and renewal risk surface before you sign.
• Services Pricing. Tracks how competing agencies and consultancies package and price their plans, so your rate card holds up against the evidence.
• Marketing Spend. Watches pacing, cost per acquisition and return on ad spend across the platforms you connect, so waste surfaces before it burns the budget.

Two gateways. One pipeline. One defensible call.

A price recommendation is only worth acting on if you can see where it came from. So every recommendation in MarginTide takes the same path.

1. Research runs go out in parallel. A research gateway sends them across a bench of licensed search and data providers. When one is blocked or slow, the next provider takes over.
2. The Extractor reads each page. Structured data first, AI only when a page is ambiguous.
3. The Composer builds the recommendation. It works from the evidence, and it never cites what it did not see.
4. The Quality Grader scores it. A five-point rubric, on a different AI vendor where possible.
5. You decide. The Action Center shows a graded call with its evidence attached. Accept it, dismiss it, or ask for more evidence.

The models come from Google, OpenAI and Anthropic and sit behind one gateway with automatic failover, and every call is logged by the model that made it. No single data source or AI vendor can let you down mid-run.

What it will never do.

Some lines are built into MarginTide, not left to settings.

• Never change a price on its own. You approve each move, or you write the rules that do.
• Never cite a page it did not open. Every figure opens to the source it came from.
• Never tie you to one AI vendor. Models and data sources are swappable parts, mid-run.
• Never hide its working. Each report carries its grade and the reasons for it.

This is the pattern we use for every agent we ship for clients: evidence you can open, and a person on every call that matters. MarginTide packages it for pricing.

Works with the stack you already run.

Accepted price moves can be written back to Shopify, WooCommerce, BigCommerce and Salesforce Commerce Cloud, or to your own signed webhook. Alerts arrive in Slack and by email when something moves that matters, not every time anything moves. Marketing Spend connects to Google Ads, Meta, LinkedIn, Microsoft Advertising, Amazon Ads and TikTok Ads.

Your own tools and your AI assistant can work in a workspace too, through the REST API, the MCP server and signed webhooks, under the same permissions and audit trail as your team.

See it on sample data.

The fastest way to judge MarginTide is to open the demo: real product screens on sample data, and no account needed. When you want it watching your own prices, Basic is free for 30 days with no card required; after that, runs need a paid plan. The MarginTide page on this site has the plans and what each one includes.