webserver-mcp
Enables AI agents to edit and serve a static website via natural language, providing file management tools over MCP and HTTP hosting.
README
webserver-mcp
Simple Python MCP server that lets agents edit a hosted website with natural
language. It exposes file tools over MCP and serves ./files/ over HTTP.
All file operations are sandboxed to <cwd>/files/ — paths are sanitized and
any attempt to escape (.., absolute paths) is rejected.
MCP tools
| Tool | Purpose |
|---|---|
file_tree(name_query="") |
List files under ./files, optional name filter |
read_file(path, start_line=0, end_line=0) |
Read full file or a line range |
search_content(pattern, path="") |
Regex search, returns relpath:lineno:line |
write_file(path, content, start_line=0, end_line=0) |
Full write or line-range replace/insert (creates file) |
delete_file(path) |
Delete file or dir |
move_file(src, dst) |
Move/rename |
Run
The server always runs as one HTTP app (uvicorn) serving both:
/— the public static site from./files/(no auth, end users view it)/mcp— the OAuth-protected MCP admin channel agents use to edit it
Auth (OAuth 2.0)
/mcp is guarded by a built-in, in-memory OAuth 2.0 provider so clients that
require OAuth (e.g. Claude.ai web) can connect. It implements metadata
discovery (RFC 8414 / 9728), Dynamic Client Registration (RFC 7591), and the
authorization-code grant with PKCE. Login is a single shared password
(MCP_PASSWORD) entered on a small web form. State is in memory, so a restart
just forces clients to re-authorize.
Run directly
pip install .
export MCP_PASSWORD=<shared-login-password> # required, server exits without it
export PUBLIC_URL=https://mcp.example.com # https origin clients reach (no trailing /)
export HOST=0.0.0.0 PORT=8000 # optional, defaults 127.0.0.1:8000
python server.py
Run with Docker
docker build -t webserver-mcp .
docker run -p 8000:8000 \
-e MCP_PASSWORD=<shared-login-password> \
-e PUBLIC_URL=https://mcp.example.com \
-v "$PWD/files:/app/files" \
webserver-mcp
The ./files volume persists site edits across container restarts.
Remote MCP client config (streamable-http)
Point the client at the /mcp URL with no Authorization header — it will
run the OAuth flow itself and open the login form in a browser:
{
"mcpServers": {
"webserver-mcp": {
"type": "streamable-http",
"url": "https://mcp.example.com/mcp"
}
}
}
For Claude.ai web: add it as a custom connector with the same /mcp URL.
Config via env:
| Var | Default | Meaning |
|---|---|---|
MCP_PASSWORD |
— | shared OAuth login password, required (server exits without it) |
PUBLIC_URL |
derived from request | public https origin; required behind TLS/a proxy so OAuth metadata advertises correct URLs and the host passes DNS-rebinding checks |
ACCESS_TOKEN_TTL |
86400 |
access-token lifetime (seconds) |
HOST |
127.0.0.1 |
bind address |
PORT |
8000 |
bind port |
Production: terminate TLS in front (nginx/Caddy) and set PUBLIC_URL to the
public https URL. The password is compared with hmac.compare_digest
(constant-time). Tokens live only in memory.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.