Your repricer now answers questions in ChatGPT and Claude

Flashpricer's new MCP server lets your AI assistant pull live repricing data, run Buy Box scans and push bulk price changes from a chat window.

Most of the work in repricing isn't repricing. It's the checking.

You open the dashboard to see which items are sitting at their floor. You filter for negative profit. You spot a SKU that looks wrong, so you open its history to figure out when it started sliding. You check the competition on it. Then you go find the strategy it's assigned to, look at the rules, decide the floor is a dollar too low, and go fix it. Then you do that again for the next one.

None of those steps are hard. There are just a lot of them, and they're spread across five screens.

Flashpricer now has an MCP server, which collapses most of that into a conversation. You connect your account to whichever AI assistant you already use, and then you ask it things. It pulls the answers live from your account.

What MCP actually is

Model Context Protocol is an open standard that lets an AI assistant connect to an outside system and use it directly. It's supported by ChatGPT, Claude, Gemini, Copilot and Cursor, plus the CLI and editor tools like Claude Code, Gemini CLI and Codex.

The practical version: instead of copying a CSV export into a chat window and hoping the model understands it, the assistant queries your live Flashpricer account and answers from real data. No API keys, nothing to paste into a config file. You go to Reports → MCP Integration in Flashpricer, pick your assistant, and authorize with your normal login. It takes about two minutes.

A fresh connection gets read access to products and orders and nothing else. If you want it to be able to change things, you approve that separately when you connect.

What this looks like on a Tuesday morning

Here are things sellers are doing with it now.

Triage without filters

"Show me every FBA item at minimum price with negative profit, and tell me which strategy each one is on." That's a catalog search, a profit filter and a strategy cross-reference in one sentence. You don't build the filter set. You describe the outcome.

Chase down a specific problem

You notice an ASIN isn't converting. "Why has this item's price moved over the last 30 days, and who's competing on it right now?" The assistant pulls the repricing history with the orders around each price change, then pulls every competing offer sorted by total price. You get the answer and the context in one pass instead of two tabs.

Find the listings you forgot about

"Which of my listings have repricing disabled?" This is the one nobody remembers to check. Items with repricing off are hidden from every other view in the system, which means a listing you paused six months ago has been sitting at a stale price ever since. One question surfaces all of them.

Bulk changes described in plain English

"Raise the min by a dollar on these 40 SKUs and turn repricing on." The assistant builds the update file and shows you a preview first: how many rows, which columns, a sample of what's about to happen. Nothing gets pushed until you say go. If a few rows fail because a SKU isn't found, those get logged in an error report and the rest still apply. Every file it pushes is stored, so a bulk change is always traceable back to what was changed and when.

Tune a strategy by describing the change

"Make my Aggressive FBA strategy a bit less aggressive, raise the floor by a dollar." Before it touches anything, the assistant pulls the current configuration so you both see the before and after. It also checks what your plan supports first, so it won't propose something that would get rejected.

Reports that stop being a chore

"Export all my negative-profit FBA items to CSV, then tell me what's in it." It creates the export, waits for it to finish, opens the file and summarizes it. If it's something you want every week, say so and it sets up the recurring version.

Buy Box by state

This one is worth calling out. Amazon's API tells you that you own the Buy Box. It doesn't tell you that you own it in 31 states and lose it in Texas. Ask for a per-state scan on an ASIN and Flashpricer goes and checks the live winner in each state. It costs one credit per state and scans all 51 by default, so the assistant shows you the estimated cost and your current balance and waits for your approval before spending anything. You can also just scan the three states you actually care about.

Walmart sellers get the same thing

Everything above works across your Walmart channels too, which matters because Walmart is where most repricers still make you do the manual work.

Ask about your Walmart catalog the same way you'd ask about Amazon. Update referral fee percentages and shipping overrides in bulk. Assign a Walmart strategy to a set of items. Pull revenue, profit, fees, COGS and units for one item or aggregated across a search.

If you run both marketplaces, this is the part that saves the most time. "Compare units sold and profit across my Amazon and Walmart channels last month" is one question instead of two exports and a spreadsheet.

The part I'd want to know about before connecting anything

Fair question: what stops this from doing something stupid to my prices?

A few things.

The connection only ever sees your organization. It's scoped to your account at the authorization layer, not by convention.

Read-only is the default, and it's genuinely enough for most of what's above. Questions, reporting, diagnostics, competition checks. If you never want the assistant to change anything, connect it with view permissions and the write tools simply don't appear in its list. They aren't there to be talked into.

Where you do grant write access, the risky operations preview first. Editing an item's overrides, pushing a manual price, re-enabling repricing on a disabled item, pushing a bulk update, creating or deleting a scheduled report: all of those show you what's about to happen and wait for your yes. A couple of operations apply directly with no confirm step, mainly assigning a strategy to an item and editing a strategy's rules, so the honest advice is to ask to see the plan before you agree to it.

Guardrails still apply. A manual price has to fall inside the item's min and max. Items with no min, no max or repricing turned off get refused. Every write lands in the item's history tagged to the assistant that made it, so you can always see what changed.

And access is a token, not a password. Revoke it from the connections list and it's dead immediately.

Why this matters more for repricing than for most software

Repricing is a system you tune, not a system you set. The margin between a good month and a flat one usually lives in a handful of listings with a floor set too low, a strategy that's slightly wrong for the category, or competition that shifted three weeks ago and nobody noticed.

The reason those go unfixed isn't that sellers don't know how. It's that finding them takes twenty minutes of clicking, and there's always something more urgent.

Cutting that to a question you ask while your coffee is still hot changes how often it actually gets done. That's the whole pitch.

Try it

If you're already on Flashpricer, this is live in your account right now. Reports → MCP Integration, pick your assistant, authorize. Then ask it something small, like "show me my products at minimum price," and see how it feels.

If you're not on Flashpricer yet, the 14-day free trial gets you the full thing including this. Setup runs about fifteen minutes, and connecting an assistant to it takes another two.

Questions about any of it, we're at success@flashpricer.com.