WebCodex MCP
An MCP server on your machine that lets ChatGPT and Claude edit code, run tests and check Git in a local repository
- 2,453 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- A cloud assistant reads files, edits code and runs commands with the Runner account permissions on your machine
- share opens a public HTTPS address guarded only by a temporary credential, and the key-in-URL mode raises the chance of a leak through logs
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/webcodexDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- A cloud assistant reads files, edits code and runs commands with the Runner account permissions on your machine
- share opens a public HTTPS address guarded only by a temporary credential, and the key-in-URL mode raises the chance of a leak through logs
This is third-party code. Review the repository files before installing.
What it does
WebCodex turns your computer into a working environment for a cloud AI assistant. The Server accepts client requests over MCP, checks access and forwards them to a Runner on the machine that holds the repository. The assistant reads and searches files, makes checked edits, runs tests, compilers and formatters, and inspects Git status and diffs. For a trial, one command in the repository folder is enough: `npx --yes @yyjeqhc/webcodex share`. It brings up a temporary HTTPS MCP endpoint with a one-off credential that stops working when the command exits. For daily use, you install a Server and a Runner on one or several machines, register projects and connect the client with the printed address. Long tasks run as observable jobs, and the model reply need not wait for them to finish. For human review there is a browser console with projects, sessions, jobs and diffs.
Who it is for. For developers who use ChatGPT or Claude in the browser and want to give them access to a project on their own machine.
Good fit when
- You want to connect ChatGPT or Claude as a custom connector to a local repository so it can run tests and check Git
- Your repositories live on several machines and you want a single Server with a Runner on each
- You want to try the idea on one project quickly with share and no permanent install
- Tasks run longer than one model reply and you need jobs and sessions you can watch in a console
Not a fit when
- The code must not be shown to an outside AI service: files read and command output go to the assistant client
- You need a sandbox: processes run with the Runner account permissions and project roots do not replace isolation
- A local agent such as Claude Code or Codex on the same machine is enough, with no intermediate server
Example request
Inspect the repository through WebCodex, summarize its structure and change nothing. Then fix one small issue, run the relevant tests and show me the diffMCP
- Transport
- http
- Authentication
- API key
Limitations
- The share command needs Node.js 18 or newer, Git and a repository you are comfortable showing to an AI. It works on Linux, macOS and Windows x64; the Cloudflare tunnel on Windows ARM64 needs your own cloudflared binary.
- Capabilities depend on the ChatGPT plan, Developer Mode and administrator policy: what the client does not allow, WebCodex cannot enable.
- Unified installers for Windows, macOS and Debian or Ubuntu still await verification on real machines.
- The license is Apache-2.0.
How to disable. In trial mode press Ctrl-C and the address and credential stop working. For a permanent install, remove the connection from the client settings, stop the Runner and unregister projects in the console or CLI.
FAQ
How do I connect WebCodex to ChatGPT?
Run `npx --yes @yyjeqhc/webcodex share` in the repository folder, wait for the WebCodex ready line and create an app in ChatGPT settings with the printed MCP URL. Choose token auth, paste the temporary credential and run Scan Tools.
How do I connect WebCodex to Claude?
Add a custom connector and paste the same printed `/mcp` address, choosing the auth method from what the CLI printed. If the client cannot send a Bearer token, `webcodex share --auth query-token` prints a URL with a temporary key in the query.
Does my code stay on my machine?
Yes, the repository is not copied into someone else's workspace, and commands run through the Runner on your side. The requested file contents and tool output still go to the assistant client, which matters for private code.
How is share different from a permanent install?
share is temporary: one project, and the public HTTPS address and credential live while the command runs. A permanent install uses a Server and Runner on your machines, registers projects and suits daily work.
Can I connect several machines to WebCodex?
Yes. One Server accepts several Runners, each on its own machine, and projects are registered on each of them. The assistant client connects to the single Server address.
Discussion
No questions or reviews yet.
Related
MCP serversA self-hosted MCP server that gives ChatGPT and other clients access to code on your machine: read, edit, search and run commands
CodexPro
codexpro
A local MCP server that gives ChatGPT in developer mode tools for reading, searching, editing and checking the repositories you allow
An MCP server that gives Claude Desktop and other clients file tools, a shell and sub-agents in the style of Claude Code
An MCP server that lets an agent read files, search code and run commands in a workspace
An MCP server and CLI for code research: search and understand your project and all of GitHub with fewer tokens
A Claude Code plugin that hands a task to a local no_human and reviews a branch with a reviewer that never saw the coder's session
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| WebCodex (this entry) | yes | API key | 2,453 | |
| AgentDock | no | API key | 1,285 | an MCP runtime through which an agent operates local machines, servers and containers, without Server and Runner routing |
| SSH MCP | no | API key | 840 | an SSH server with command classification and policies for remote hosts, with no repository workspace for Git and jobs |
README in short
The idea is to give a cloud assistant a real development environment on your machine through MCP. A diagram traces the path from the client through Server to a Runner next to the repository, and separate documents cover the console, session continuity, deployment and troubleshooting.