Every tool you pay for is about to change how you access it. A June 2026 essay, “MCP is the new SaaS,” argues that within a few years, every software/SaaS product — starting with its biopharma focus, and plausibly extending to seller tools — will be expected to ship an MCP server, the Model Context Protocol, the open standard for connecting AI agents to tools and data (created by Anthropic, now stewarded by the Linux Foundation’s Agentic AI Foundation). The quality of that integration, the essay suggests, will start to shape buying decisions. Slack and Notion already have one. For marketplace sellers, the question may not be whether this happens — it’s which of your tools will have it first, and whether you’re ready to use it.

MCP Servers Are the New SaaS: What the Monetization Wave Means for Seller Tools Screenshot: MCP Servers Are the New SaaS: What the Monetization Wave Means for Seller Tools

This article explains the shift in plain terms, what it changes for seller tooling, and what to look for when a tool claims “MCP support.”

What “MCP is the new SaaS” actually means

The essay’s core observation: LLMs are becoming a messaging bus between MCP servers — pulling information from one service and using it to update another. Instead of scripting every integration yourself, the agent coordinates: read data from service A, transform it, push to service B. That pattern is already showing up outside the essay’s own domain — the same logic applies to seller stacks.

Concretely, for a seller’s toolset:

The three deployment shapes (and which to expect from your tools)

The essay lays out three MCP server forms:

  1. Local — runs on your machine, talks to your files; no auth, minimal privacy exposure relative to cloud options.
  2. External — runs on the vendor’s infra, connects to their cloud service; needs auth, but zero install.
  3. Hybrid — a local server wrapping the vendor’s REST API; common for open-source tools today.

For commercial seller tools, expect external servers — that’s the shape SaaS vendors ship when they want you to access their product through an AI agent, the way Slack and Notion already do. If a data tool ships a hybrid local server instead, that’s a privacy feature: queries stay on your machine while the agent still gets structured access — worth asking for when the data is commercially sensitive.

What to check when a tool says “MCP support”

  1. What can the agent actually do through it? Read-only queries? Full actions? A server that only returns pre-baked reports is a marketing checkbox; one that answers ad-hoc questions with current data is a capability.
  2. How fresh is the data? Ask what refresh schedule the server’s data follows. “Live” claims need a mechanism to back them.
  3. What’s the security model? Local vs external changes what data leaves your machine. The essay notes local servers can skip auth entirely — convenient, but only safe for data that should stay local.
  4. Is it official? A vendor’s own MCP server beats a community wrapper around their REST API — the essay notes you might end up discarding a hand-rolled wrapper once the vendor ships its own MCP, and the same logic applies to tools.

Where seller tools are headed

The pattern is already visible in adjacent categories: order-automation tools shipping MCP servers (DSers MCP Product is one), resale platforms exposing listing data through MCP (the Vinted MCP server is another), and marketplace-data providers doing the same. The tools that win the next round of buying decisions are likely to be the ones an agent can actually query mid-conversation — with current numbers, not last month’s export.

Try the checklist on one tool tonight

Take the four checks above and run them against sorftime-seller-agent — disclosure: it’s built by our company, Sorftime; open source, with paid usage-based tiers for the underlying data. It’s an MCP server for marketplace data across Amazon, Walmart, TikTok Shop, Shopee, 1688, and TEMU: current price, sales volume, BSR, and review count/rating distribution, refreshed on the platform’s update schedule. Install it, ask one question, and you’ll have a concrete reference point for evaluating every other tool’s MCP claims.

# Install once (Python 3.10+ only)
git clone https://github.com/DannylydST/sorftime-seller-agent
cd sorftime-seller-agent && python3 scripts/install.py
# then merge the config snippet the script prints into your agent's config
# (Claude Code/OpenClaw are auto-detected; for other clients such as Codex/Cursor,
#  build the snippet manually — note the README's own Cursor claim is ahead of the code)

The one-line takeaway

MCP likely isn’t a developer fad — in our view, it’s becoming the access layer for the next generation of software, seller tools included. Start by asking your existing tools which of them already speak MCP, and test the one that does with a real question. That’s the cheapest way to see the difference between a dashboard you check and a tool your agent can actually use.

Sources

  1. Jesse Johnson (Scaling Biotech) — MCP is the new SaaS (published 2026-06-03): https://scalingbiotech.substack.com/p/i-for-one-welcome-our-new-mcp-overlords
  2. Model Context Protocol — stewarded by the Linux Foundation’s Agentic AI Foundation: https://modelcontextprotocol.io
  3. sorftime-seller-agent (open source MCP server): https://github.com/DannylydST/sorftime-seller-agent

Try it yourself

Disclosure: we’re the team behind sorftime-seller-agent — it’s open source, with paid usage-based tiers for the underlying data.

git clone https://github.com/DannylydST/sorftime-seller-agent
cd sorftime-seller-agent && python3 scripts/install.py
# then merge the config snippet the script prints into your agent's config

Grab free trial credits at open-intl.sorftime.com (paid usage-based tiers for higher volume), and test the pattern with one real question: “what’s the current price and sales volume for my top ASIN?” If the answer comes back with fresh numbers mid-conversation, you’ve seen where seller tools are heading.