For agents
Connect your agent to this site
Everything on this site about Leevon Levasseur is open to your agent: over MCP, with three tools, or over a plain HTTP API. No key, no sign-in. Copy the block for your client and ask away.
The MCP server
The server speaks the Model Context Protocol over Streamable HTTP at the URL below. It is stateless, so every request stands on its own and no session is issued.
https://leevonlevasseur.com/mcp| Tool | Input | Returns |
|---|---|---|
search_kb | query, and an optional limit (1 to 50, default 5) | The best-matching sections of the site, each with a link to its heading. |
get_page | path, such as /resume | One page in full, section by section. |
ask | question, up to 2,000 characters | The guide's answer, with the pages it drew on as citations. |
Every page in the knowledge base is also an MCP resource, so a client that lists resources sees the site's pages by URL and can read any of them in full.
Connect a client
Each block below is complete. Paste it, and the server shows up as personal-site with its three tools.
claude mcp add --transport http personal-site https://leevonlevasseur.com/mcpSettings → Connectors → Add custom connector
Name: personal-site
URL: https://leevonlevasseur.com/mcp{
"mcpServers": {
"personal-site": {
"url": "https://leevonlevasseur.com/mcp"
}
}
}{
"mcpServers": {
"personal-site": {
"type": "http",
"url": "https://leevonlevasseur.com/mcp"
}
}
}{
"mcpServers": {
"personal-site": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://leevonlevasseur.com/mcp"
]
}
}
}curl -sS https://leevonlevasseur.com/mcp \
-H 'Content-Type: application/json' \
-H 'Accept: application/json, text/event-stream' \
-H 'MCP-Protocol-Version: 2025-06-18' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'The curl request answers in a text/event-stream body: one data: line holding the JSON-RPC result with search_kb, get_page and ask. No initialize call is needed first, because the server keeps no session.
Limits
- Ten asks a minute. Each IP gets ten
askcalls per minute, shared withPOST /api/v1/ask. Past that, the call fails with 429 andRetry-After: 60. - One daily model budget. The whole site shares one cap on model calls per UTC day. Once it is spent,
askstill answers, with no error and no citations, and the answer is:I'm resting until tomorrow. I've answered as many questions as I can today, so ask me again after midnight UTC.
search_kbandget_pagecall no model and keep working. - Would rather use HTTP? The same data is at
/api/v1, described by the OpenAPI document at /api/v1/openapi.json and rendered below.