MCP and Skills — concepts and AIEO's plan

Two integration concepts keep coming up in discussions about AI-assisted workflows: MCP and Skills. This guide explains both at a conceptual level, sketches how AIEO plans to use them, and is explicit about what does not exist today.

What MCP isCopy Anchor Link

MCP (Model Context Protocol) is an open protocol that lets AI clients — chat apps, IDEs, agents — talk to external systems through a server. A server exposes tools: typed operations a client can call, such as "read a report", "list issues", or "fetch a page's evidence".

The value for a product like AIEO is read-style access to structured data. Instead of copying numbers out of a browser, an MCP-compatible client could ask:

• "List the open technical findings in my latest AIEO report."
• "Show the evidence for finding CRAWL-RENDER-001."
• "Which findings changed between the last two reports?"

The assistant stays the interface; the product stays the source of truth.

What Skills areCopy Anchor Link

Skills are reusable, written workflows that an agent can load and follow: procedures with steps, guardrails, and definitions of done. Where MCP is about access to data and operations, Skills are about procedural knowledge — how to do a class of job the same reliable way every time.

A useful mental model:

• MCP = "the agent can read my reports and issue data."
• Skills = "the agent knows the house process for turning a finding into a verified fix."

How AIEO plans to use themCopy Anchor Link

AIEO's planned agent integration follows one draft flow:

1. Read the report and its summary.
2. Fetch one finding with its evidence — the observed value, location, and rule.
3. Draft a change plan scoped to that finding.
4. Wait for human confirmation. This step is required, by design — suggestions never auto-apply.
5. Apply and re-verify: after the change, a fresh check compares against the recorded evidence.

The draft flow exists to keep one property intact: the agent accelerates the loop described in AIEO, GEO and SEO — understand, evidence, improve, verify — without ever removing the human from the decision.

What does not exist yetCopy Anchor Link

To be precise about the current state:

• No MCP server URL — AIEO does not operate an MCP endpoint today.
• No npm package or install command — nothing has been published under the AIEO name.
• No compatibility claims — no statement is made here about which AI clients will or will not work with a future interface.

When any of this ships, the developers page will document the real endpoint and configuration — not before.

What MCP/Skills will NOT doCopy Anchor Link

Two misconceptions worth pre-empting:

• MCP support is not an SEO or AIEO ranking factor. Whether your site uses agent interfaces has no documented effect on how search or answer systems rank or cite content.
• Agent workflows are not autonomous optimization. The planned flows end at human-confirmed changes, because unreviewed automated changes to a live site are how small mistakes become outages.