For agents and the people who run them

Connect your agent

This site exposes real, typed tools to AI agents — not a chat widget, and not a page an agent has to drive by clicking. Point your coding agent at one URL and it can scan any website for accessibility failures, pull exact pricing, and search and read the whole guide library.

The endpoint

no key required

https://fullstacktechnyc.com/api/mcp

Streamable HTTP MCP, protocol 2025-06-18, server version 1.0.0. The 6 public tools below are read-only and anonymous — nothing there can spend money, send mail, or change stored state, so there is nothing to authorise and no token to manage. Workspace customers get 6 more against their own data; see below.

Connect

pick your client

Claude Code

One command, from any directory:

claude mcp add --transport http fullstacktechnyc https://fullstacktechnyc.com/api/mcp

Then ask it to scan a site. Run `/mcp` to confirm the tools loaded.

Claude Desktop / Cursor / Windsurf

Add to the MCP config file (Settings → MCP, or claude_desktop_config.json):

{
  "mcpServers": {
    "fullstacktechnyc": {
      "type": "http",
      "url": "https://fullstacktechnyc.com/api/mcp"
    }
  }
}

Restart the app after editing. Cursor and Windsurf use the same shape.

VS Code (GitHub Copilot)

Add to .vscode/mcp.json in the workspace:

{
  "servers": {
    "fullstacktechnyc": {
      "type": "http",
      "url": "https://fullstacktechnyc.com/api/mcp"
    }
  }
}

Then pick the tools from the Agent-mode tool picker.

Anything else

Any MCP client that speaks Streamable HTTP. No SDK, no key:

curl -X POST https://fullstacktechnyc.com/api/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

No authentication is required for the public tools.

The tools

6 available

Tool	What it does	Input
`fst_scan_accessibility`	Run a real accessibility scan against any public website's homepage and return a 1-5 score plus a per-rule breakdown of violations.	`url` string · required
`fst_get_pricing`	Return exact, current pricing for Full Stack Tech NYC's school and PTA website service, including what is included, what the previous vendor charged, ownership terms, and the free tier for Title I schools.	`option` string
`fst_list_services`	List the services Full Stack Tech NYC offers — school and PTA websites, web apps, AI and LLM integration, MCP and agent tooling, Cloudflare edge infrastructure, and accessibility remediation — with who each is for.	none
`fst_search_guides`	Search Full Stack Tech NYC's published guides and articles by keyword and return matching titles, URLs, and their answer-first summaries.	`query` string · required `limit` integer
`fst_get_guide`	Fetch the full text of one published guide or article by its slug, as returned by fst_search_guides.	`slug` string · required
`fst_start_project`	Return the right link and the information worth having ready to start a project or book a free 20-minute call with Full Stack Tech NYC.	`kind` string

Your own workspace

6 tools · needs a token

If you have a Full Stack Tech workspace, the same endpoint gives your agent access to your data — search your uploaded documents, read past conversations, check your balance, and ask questions grounded in your own material. Create a token at Settings → Connect agent and send it as a bearer token:

claude mcp add --transport http fullstacktechnyc https://fullstacktechnyc.com/api/mcp \
  --header "Authorization: Bearer YOUR_TOKEN"

Tool	What it does	Scope
`fst_account_status`	Show the signed-in user's Full Stack Tech workspace: team name, credits remaining, chat messages remaining, and how many documents are uploaded.	read
`fst_list_documents`	List the documents uploaded to the user's Full Stack Tech workspace, with their processing status.	read
`fst_search_documents`	Semantic search across the documents in the user's Full Stack Tech workspace.	read
`fst_list_conversations`	List the user's saved conversations in their Full Stack Tech workspace, newest first.	read
`fst_get_conversation`	Read the full message history of one saved conversation by its id, as listed by fst_list_conversations.	read
`fst_ask_workspace`	Ask a question answered by Full Stack Tech's assistant, grounded in the documents in the user's workspace, and save it to a conversation.	write · uses credits

Tokens are scoped. A read token can never spend your credits, which matters more than it sounds — an agent loop that misfires against a read token wastes time, and against a write token wastes money.

In the browser

WebMCP

The same tools are registered with in-browser agents through WebMCP, the document.modelContext API in the Chrome 149–156 origin trial. If your browser supports it, an agent running inside the page can call these tools directly instead of guessing at the DOM.

To try it today: enable chrome://flags/#enable-webmcp-testing and the DevTools WebMCP flag, which adds a panel listing every registered tool and lets you invoke one by hand. Note the API lives on document, not navigator — a lot of writing from early 2026 says otherwise and is describing a superseded draft.

Why bother

the argument

Most "AI-ready" websites are ordinary websites with a chatbot bolted on. An agent sent to do something real still has to read rendered HTML, guess which button is the submit button, and hope the markup did not change. That is screen-scraping with extra steps, and it breaks constantly.

Exposing typed tools inverts it: the site declares what it can do, the agent picks a function and passes arguments, and the contract is explicit enough to version. The accessibility scanner above is the honest test of that claim — it is the same code path the website form uses, reachable by an agent in one call, with no browser automation anywhere in the loop.

We build this for clients too. If you want your product callable by agents — WebMCP, a remote MCP server, or both off one registry, which is how this site does it — get in touch.