Developers
Two read-only doors into the same public facts: an HTTP JSON API and a stateless MCP server. Nothing here stores a message. Payments are off.
HTTP API
Responses are JSON. Errors are JSON too. CORS allows any origin on these routes. Only GET is accepted.
Profile
curl -s https://keithmitchell.dev/api/v1/profile Experience
curl -s https://keithmitchell.dev/api/v1/experience Wrong method
curl -i -X POST https://keithmitchell.dev/api/v1/profile That returns HTTP 405 and a body shaped like {"error":{"code":"method_not_allowed","message":"Use GET or HEAD."}}.
The contact email in the payload is [email protected]. Keith still needs to confirm Cloudflare Email Routing for it.
MCP
https://keithmitchell.dev/mcp speaks Streamable HTTP, stateless, using the official TypeScript SDK.
Each request builds a fresh server. There is no session id. Tools advertise readOnlyHint and reject unknown input keys.
get_profile— name, email, summary, GitHub links, knowsAboutget_experience— resume roles plus labeled public repositoriessearch_skills—{ "query": "ruby" }get_contact— email and GitHub URLs
Client config
{
"mcpServers": {
"keith-mitchell": {
"url": "https://keithmitchell.dev/mcp"
}
}
} MCP Inspector
Against the public site, after it is deployed:
npx --yes @modelcontextprotocol/inspector --cli https://keithmitchell.dev/mcp --method tools/list Against the local dev server:
npx --yes @modelcontextprotocol/inspector --cli http://127.0.0.1:44731/mcp --method tools/list A script in this repo does the same check with the SDK client: npm run test:mcp.
For agents
- /llms.txt
- /llms-full.txt
- /resume.md
- /robots.txt allows AI crawlers
- /sitemap.xml includes lastmod