About krexel-mcp
krexel-mcp is Krexel’s MCP server for deploying and editing static sites through MCP-compatible AI clients. Its tools cover site files, deployment history, rollbacks, environment variables, and status.
Key features
- 01Deploy built site folders
- 02Edit deployed files
- 03Inspect and read site files
- 04List deploys and view logs
- 05Rollback to a previous deploy
- 06Manage environment variables and status
Use cases
- 01Publish a built static site from an AI client
- 02Ask an AI to inspect and revise deployed files
- 03Restore a previous deployment after a change
MCP support
Exposes nine Krexel tools to MCP clients through standard tool discovery. The tools support deployments, file operations, logs, rollback, environment variables, and account status.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
9Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| domain | string | Yes | Custom domain to attach (e.g. shop.example.com). |
| folder | string | Yes | Absolute or ~-relative path to the built site directory. |
| framework | enum | No | Build framework hint. Default 'auto' detects from package.json.auto | nextjs | astro | vite | static |
Details
From the project READMEkrexel-mcp
Ship AND edit static sites from any AI assistant that speaks the Model Context Protocol.
krexel-mcp is the official MCP server for
Krexel — a "your AI ships and edits your live site"
API. Point any MCP-compatible client (Claude Desktop, Cursor, Windsurf,
VS Code) at this server and your assistant can deploy, list, edit,
rollback, and inspect sites just by talking.
Quickstart
Add krexel-mcp to your MCP client configuration:
{
"mcpServers": {
"krexel": {
"command": "npx",
"args": ["-y", "krexel-mcp"],
"env": {
"KREXEL_API_URL": "https://api.krexel.com",
"KREXEL_API_KEY": "krx_live_••••••••••••••••",
"KREXEL_MASTER_KEY": "<your-32-byte-hex-key>"
}
}
}
}
Restart your client. The assistant now sees nine Krexel tools.
Tools
| Tool | What it does |
|---|---|
ship_site | Upload a built site folder. Returns a deploy_id and preview URL. |
edit_file | Patch a deployed site (create / replace / replace_all / delete). 0.1 deploy. |
list_files | List the files in a deployed site (path + size + sha256). |
read_file | Read one file's content from a deployed site. |
list_deploys | List recent deploys from local state. Optional domain filter. |
get_logs | Fetch build logs for a deploy. |
rollback | Mark a previous deploy as the current good version. |
set_env | Encrypt and persist an env var for a domain (AES-256-GCM). |
get_status | Return account, deploy count, plan, and state directory. |
The conversational-edit flow
- AI calls
list_files(deploy_id)to see what files exist on the site. - AI calls
read_file(deploy_id, path)to read the file it wants to change. - AI generates the new content itself, then calls
edit_file(deploy_id, op: "replace", file, find, value, message). - Krexel ships the patch as a new deploy — live in ~8 seconds, 0.1 deploy
against the customer's quota. The new
deploy_idandparent_deploy_idare returned so the AI can chain further edits on top.
The AI is the brain. Krexel is the hands.
All tools take plain JSON arguments and return JSON. Tool schemas are
discoverable via the standard MCP tools/list request.
Configuration
| Env var | Required | Default | Notes |
|---|---|---|---|
KREXEL_API_URL | no | https://api.krexel.com | Base URL of the Krexel orchestrator API. |
KREXEL_API_KEY | no* | ~/.krexel/auth.json | Account API key. Forwarded as Authorization: Bearer *** to every API call. *Required unless saved by krexel login. |
KREXEL_MASTER_KEY | yes* | — | 32-byte hex string for env-var encryption. *Required for set_env. |
State files are written under $KREXEL_HOME (default ~/.krexel/).
License
MIT © Syfer
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.