Why I built it
I kept hitting the same wall. I'd be mid-flow in Cursor, or thinking through a pattern on my phone, and want a quick, trustworthy answer on WCAG or an assistive tech question. Switching tools, searching, sanity-checking. It broke the thread every time. I've been doing accessibility work for years (you can read more about my background and approach on the about page), but this was a problem I hadn't solved yet.
So I built Inclusive Intelligence MCP. It's a remote Model Context Protocol(opens in new tab) server that gives any AI assistant deep, curated accessibility knowledge: WCAG 2.2 in full, universal design principles, disability categories and lived barriers, assistive technology, laws like the EAA and Section 508, and inclusive language. Every answer is grounded in real human experience, not just spec text. If you're aiming for WCAG 2.2 AA compliance, having this knowledge available on demand means fewer skipped criteria and faster remediation cycles.
What it covers
The MCP is built around the work I do most: accessibility audits and reviews. I ask an AI to review a design or a production app, and the MCP tells it exactly what to look for with each WCAG criterion: no mixing things up, no guesswork. The same goes for implementation reviews of HTML, CSS, React, and mobile UI. It can even guide an AI to write accessible code, so even smaller models produce solid results as long as they have enough context.
It also covers universal design validation, global legal frameworks (ADA, Section 508, EN 301 549), disability etiquette, and inclusive language. Every response is grounded in WCAG 2.2 AA criteria, with AAA guidance where applicable. Built-in report templates, including VPAT, mean every audit result and every report follows the same structure, so your output is consistent whether you're working alone or handing off to a team.
Why remote matters
The reason it's remote is this: one config and it works from Claude, Cursor, Codex, my phone, anything that speaks MCP. I'm not tied to a single editor and I'm not copy-pasting the same context into five different tools. The AI just knows where to look.
The MCP server configuration is minimal. Add a few lines to your project and you're set:
{
"mcpServers": {
"inclusive-intelligence": {
"url": "https://mcp.moreinclusive.se/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Place that in .cursor/mcp.json for Cursor, or .mcp.json for Claude Code and other clients.
How it changes the conversation
What I like most is the shift in quality. Instead of relying on a model's general knowledge, the MCP gives it curated expert knowledge directly. My old problem was models mixing up WCAG criteria from their own training. Now the AI pulls the actual source, so it knows exactly what to look for with each criterion. Fewer hallucinations, fewer "actually the spec says" moments. It feels less like using a chatbot and more like having a well-read colleague on tap.
I use it daily. I ask the AI to review a design or a live app, and the MCP makes sure every check is grounded in the real criteria: focus order, colour contrast, status messages, touch targets, you name it. The AI handles the inspection, the MCP handles the expertise. I stay in my flow, and the output is consistent because the MCP pulls from the same source every time.
A concrete example: I hand an AI a page URL or a set of design screens and ask it to audit against WCAG 2.2 AA. No additional context, no brief, no pointing at what to check. The MCP tells the AI exactly what each criterion requires, what to look for, and how to report it. Every audit checks the same things the same way, and every report comes back in the same format: VPAT-compatible, with severity ratings and remediation steps. That consistency alone has pushed my efficiency up by several hundred percent. I'm not re-explaining the criteria to the tool; I'm just reviewing the output.
The same applies to design reviews. I ask the AI to review screens against universal design principles, and the MCP makes sure every principle is considered, not just the obvious ones. The result is a structured review every time, without me having to remember which principles apply to which surface.
The same setup also works for code. Because the MCP has implementation guidance for HTML, CSS, React, and mobile UI, even smaller models can produce accessible code with the right context. It's not just about catching problems; it's about shipping fixes too.
Get access
The server uses an API key for authentication, so you'll need to request one before you can connect. Free for personal and open-source projects: non-commercial use, side projects, community work. If you're on a commercial team, reach out and I'll set you up. Request a key on More Inclusive(opens in new tab) or reach out to me directly.
I'm still shaping what it should cover next. If you use it and something's missing, tell me. I also write about accessibility and inclusive design on this blog, and you can learn more about my broader approach on the home page.