A chatbot answers. An agent acts.
Ask ChatGPT or Claude to help with your CV today, and the normal flow is: copy your CV into the chat, describe the job, get advice back, copy the result somewhere else. Every step is manual, and the AI never actually touches your real data. It only sees whatever you pasted into that one conversation.
MCP (Model Context Protocol), an open standard Anthropic published in late 2024, changes what "an AI reads your CV" can mean. Instead of one-off copy-paste, an MCP-compatible agent connects directly to a service — in our case, your Smart CV Writer profile — over a standard interface, the same way it would connect to any other tool. No custom integration built per app, and no re-explaining your background at the start of every conversation.
What our MCP server actually does
We shipped read access first: an agent can pull your full profile, every experience, every skill, your existing Fit Checks — the same data you'd see in the app, structured for a model to actually use. Since September 2026, it can also act: add a new source to your profile, kick off a Fit Check against a job posting you paste into the conversation, re-evaluate one that's gone stale, and send feedback straight to us. Each write action runs under your own account's real permissions and its own billing gate — a Fit Check triggered from an agent conversation checks the same Active Tailoring subscription status a Fit Check triggered from the website does.
Why this matters more than it sounds like
The obvious use case is convenience: ask your agent how you stack up against a posting without opening a browser tab. The less obvious one is that your career data stops being locked inside one app's interface. Any agent that speaks MCP can act on your actual, current profile instead of a snapshot you happened to paste in once. Update a role in the app, and the next agent conversation sees the update. No re-syncing, no stale copy floating around a chat log somewhere.
What this isn't
An MCP-connected agent isn't automatically authorized to do anything you haven't approved. Connecting one requires you to explicitly authenticate the connection, the same OAuth-style flow any third-party integration goes through. And it doesn't replace the app — it's another way in, useful when you're already deep in a conversation with an agent about a job hunt and don't want to break flow to open a separate tool.
Related guides
- The structured career profile — the data an agent is actually reading
- How CV screening actually works in 2026 — what a Fit Check triggered through MCP is actually doing