Your AI is smart. It just can't see your Amazon data.

Commerce Spine is SellerPlex's managed Amazon data warehouse and API for sellers, agencies and AI builders. It stores and joins Amazon account data so your own AI agent can query it and build reports, dashboards and alerts, and its action layer applies reviewed ad-change proposals on Sponsored Products. Start with how it works, check data coverage, or compare plans and limits.

It applies reviewed proposals on Sponsored Products: keyword bids, campaign status, negative keywords and placement bids, built on the same warehouse your agent reads. Every change is logged, and you can follow each one in the console's action queue. An AI suggestion layer comes next, once organic data lands in the warehouse: it will propose what to change and explain why. Book a demo to see it working.

Because you already have one you like. Bundled AI means someone else picks the model, routes your data through it, and meters your usage in credits. Commerce Spine ships as a plugin, skills, and an API token for the harness you already run: Claude Code, Codex, Cursor, or anything that can call an API. We would rather be the best data layer than the fourth-best chat window.

Anything that can hold an API token and make an API call. The official plugin gives Claude Desktop and Claude Code first-class access: install it, sign in once, and your agent knows the full data dictionary. From Codex, Cursor, custom agents or plain scripts you use the same API directly with a scoped token, and the documented schema makes that straightforward. If your setup can make an HTTPS request, it can query your warehouse.

You can absolutely stack Amazon exports into your own database, and for a weekend project that works. The problem is not the first build, it is the second year of owning it. The Amazon API surface is genuinely complex, and the data volume needs real infrastructure: constant updates, curation and organization before an agent can use it reliably. Vibe-coded pipelines tend to end up expensive, token-heavy and inconsistent, because the hard part is not the code, it is the underlying architecture. Ours took months to tune, full time, by people who do this for a living. There is also paperwork you cannot code around: most of the deeper Amazon data requires an approved application in Amazon's marketplace. Commerce Spine already carries that approval. Then there is the part nobody budgets for: Amazon changes report formats, deprecates endpoints and shifts fields, and every one of those is an evening you spend repairing a pipeline instead of running your business. That maintenance never ends, and it is the actual cost of the DIY route. You plug in and start building; the operational headache stays ours.

Flat monthly subscription, priced by connected accounts. Free is one connected account, 30 days of backfilled history, 20 requests per hour, 40 per day, a 30-day window per query and 1.5 GB scanned per month. Paid is $49/mo for up to 3 connected accounts (brand and marketplace pairs) and $149/mo for up to 12, with custom pricing above that: 60 requests per minute, 1,000 per day, a 90-day window per query, and 150 GB scanned per month. First sync backfills 30 days on every plan, which is what Amazon releases in practice against a 90-day ceiling, and everything from that day forward is kept. Both paid plans include 5 hours with one of our senior AI developers, usable for onboarding, for migrating tooling you already built, or for standing up your first dashboards and alerts. Limits are applied per organization rather than per token, so extra keys do not buy extra quota. No credits, no overages, no per-seat fees. Your AI token costs are paid by you, directly to your model provider, exactly as they are today.

Only through the action layer, which applies reviewed proposals to your Sponsored Products ads: keyword bids, campaign status, negative keywords and placement bids. Every change is logged. The data API reads, joins and serves your Amazon data; it does not change bids, listings, prices or inventory. When you want to act on an insight yourself, you decide how: through your own tools and your own code, where you can read every line before it runs. Every request is logged with the token that made it.

As fresh as Amazon allows: performance rollups daily. Every live entity in the console shows its last successful sync, so you never have to guess whether a number is current.

Tools in that category bundle a chat AI, workflows and data behind one login and meter everything in credits, with model routing they control. Commerce Spine is the opposite architecture: we do the data layer, expose it through an AI-first API, and you bring your own AI on your own subscription. No captive UI, no credit meter, and your data only ever goes to the model you chose. If you are comparing us against a specific tool you already pay for, ask us on a call and we will walk through it honestly. Our DataHawk alternative page does that for one of them.

A warehouse, and the difference decides what your agent can actually do. A connector proxies Amazon's live endpoints: it asks the question, hands back the answer and keeps nothing, so there is no history, no joins across sources, and every comparison means re-pulling data you already fetched. We ingest from Amazon Ads and Seller Central into storage you own a slice of, clean and normalize it, join it so a campaign row can meet an inventory row, and keep it. That is why your agent can answer "compare this week against the same week last quarter, by brand" in one query instead of orchestrating a dozen calls and hoping the rate limits hold.

Amazon's Selling Partner API provides programmatic access to seller and vendor data. Building on the Amazon SP-API yourself means maintaining authorization, ingestion, retries, storage and joins across the sources you use. Commerce Spine maintains that warehouse and exposes a documented API. Its GraphQL interface lets your agent request named fields and entities. Every plan backfills 30 days at first sync, then retains data from that point forward. An MCP server is an interface to tools and data; it does not, by itself, provide this storage or maintenance. See the official Amazon SP-API overview and Commerce Spine API documentation.

No. Claude Desktop and Claude Code use the official Commerce Spine plugin. Codex, Cursor, custom agents and scripts can call the same API directly with a scoped token. MCP is a protocol for connecting AI applications to tools and data; it does not inherently require another AI model or an extra model fee. Commerce Spine's distinction is the managed warehouse and documented API underneath your agent. Read the MCP architecture overview for how the protocol works.

No, and we understand why you are asking: an agency that also sells you software is exactly where you should push. Two things keep it clean. First, isolation by architecture: every organization gets its own segregated warehouse, so there is no shared pool where your numbers could sit next to a SellerPlex client's. Second, separation by department: the agency side of SellerPlex has no path to customer warehouses. The people who can touch the infrastructure are the engineers who maintain it, for maintenance only, the same arrangement you already accept from every database vendor you use. On your side, access is limited to the people you invite through Team roles and the scoped API tokens you issue, every request is logged with the token that made it, and you can revoke any token instantly.

Your data lives in a managed data warehouse on Google Cloud, connected through Amazon's own consent flow, which you can revoke at any time. Access on your side is controlled by scoped, revocable API tokens and console team management. Your prompts and query results flow between your machine and your chosen model provider under your agreement with them. We never proxy or read your AI conversations.