Run private AI, SSH, file sharing, or GPU workloads on your own computer — and let invited users access them securely from anywhere. No public IP. No domain. No VPN. No port forwarding. No relay in the data path.
Built in C/C++. End-to-end encrypted. MTCode does not process or store application-layer data.
MTCode DirectLink is a signaling-assisted peer-to-peer platform that lets private services on your computer become securely accessible to invited users anywhere. It handles authentication, authorization, service discovery, and connection setup, while application data flows directly between endpoints over an encrypted P2P channel.
Install MTCode Server on the computer that hosts your service, then register the local TCP port you want to share. Authorized users open MTCode Portal, where the service appears automatically and is mapped to a local address on their computer — as if it were running locally.
AI servers — host chat, coding, and image-generation services using Ollama, LM Studio, vLLM, or the built-in MTCode-LLM and MTCode-Diffusion servers.
Remote GPU execution — use MTCode GPU Server and MTCode RemoteGPU to let authorized users run Python workloads on your GPU in a restricted sandbox.
AI agent gateways — publish tools such as OpenClaw by registering the single local port used by their browser-based interface.
SSH, remote desktop, and private web apps — expose only the service you choose, not the whole computer, and keep it inaccessible from the public internet.
Private collaboration tools — share Nextcloud, Gitea, Immich, or other self-hosted apps while keeping documents, repositories, and media within infrastructure you control.
Explore use-case walkthroughs: Private AI for Friends & Family • For ML Researchers • For Developers • For Artists • For Teams & Collaboration
Reverse-proxy and tunneling tools route traffic through an intermediate relay server. That extra hop can add latency, limit throughput, create bandwidth costs, and put a third-party provider in the traffic path.
Mesh VPNs connect entire devices into a private overlay network. They are useful for personal infrastructure, but when you want to share only specific services — not the entire computer — across multiple administrators, careful access-control configuration and coordination are often required.
With MTCode DirectLink, users have one app — MTCode Portal — one account, and a unified interface to access all authorized services, regardless of who operates the underlying servers. Access is granted at the individual service level, not the whole computer.
On a desktop computer with a RTX 4090 24 GB GPU, MTCode Server registers three servers: SSH Server, MTCode-LLM, MTCode-Diffusion
Register SSH Server
Register LLM Server
Register Diffusion Server
On a remote laptop computer, MTCode Portal connects to authorized servers
Get started for free on your home or office computers — no cloud hosting subscriptions or domain name registration required. The Free Personal plan covers 5 users including the administrator, with unlimited servers under that account. Plenty for a family, a small group of friends, or a personal demo. See pricing for Team, Business, and Education tiers with larger user buckets.
All registration and login traffic is protected by SSL/TLS encryption. Your servers remain on your private network behind your router, invisible to the Internet. There are no unsolicited visits or scans — only the people you invite can gain access. Only the registered servers, not the whole computer, are accessible to remote users.
All data traffic between MTCode Server and MTCode Portal is transmitted directly between the two endpoints for best performance — no intermediate server tunneling is involved. Data is fully encrypted to protect privacy.
You may already use a tunnel, a mesh VPN, or a cloud AI service. Here is what changes when you use MTCode DirectLink instead.
| Reverse-proxy / tunnel (e.g., ngrok, frp, Cloudflare Tunnel) |
Mesh VPN (e.g., Tailscale, ZeroTier) |
Cloud AI (hosted AI services) |
MTCode DirectLink | |
|---|---|---|---|---|
| What is exposed | A public URL on the Internet | Device or network-level access, depending on ACL configuration | N/A (you have no server) | Only the specific services you choose |
| Who can connect | Anyone who finds the URL | Every device on your mesh | Anyone with an account | Only invited users with access to that service |
| Public IP / domain / router config | Not required | Not required | N/A | Not required |
| Application data path | Through tunnel provider; plaintext visibility depends on TLS setup | Direct or encrypted relay fallback | Processed by hosted AI provider | Direct between endpoints; not relayed through MTCode infrastructure |
| Access services from multiple owners | Possible — each admin publishes a separate URL, no unified view | Possible, but typically requires separate setup for each owner — no unified service view by default. | N/A | One app, one account — all authorized services from all administrators in one unified interface |
Publish any service on your home or office computer and let the people you invite reach it privately from anywhere — no public IP, no domain, no port forwarding, no vendor in the data path.
Installed on the server computer, MTCode Server registers one or more local services — an AI chat server, an SSH server, a file-sharing app, or any TCP service — with the MTCode DirectLink platform and makes them reachable by authorized users anywhere on the Internet. It handles all the networking automatically: no public IP, no domain, and no router configuration required. It also ships with built-in MTCode-LLM and MTCode-Diffusion servers so you can start serving AI models immediately without any third-party software.
Installed on each user's computer, MTCode Portal retrieves all the servers the user has been granted access to — across any number of administrators — and presents them in a single unified interface. It maps each remote server to a local port so existing client software (AI chat apps, SSH clients, web browsers) connects to it exactly as if it were running locally. All traffic flows directly between the user's machine and the server computer, end-to-end encrypted, with no vendor relay in the path.
Share a GPU-equipped computer with remote users for running Python projects — securely, directly, and without a public IP. Built on the same core libraries for the MTCode DirectLink platform, each can operate independently without separate installation of MTCode Server or MTCode Portal.
Installed on a GPU-equipped computer, MTCode GPU Server lets you share GPU compute without giving users full system access. It accepts Python project submissions from authorized remote users and executes them inside a restricted sandbox — isolating each job from the host system and from other users' work. Results stream back to the user's editor in real time, or jobs can be submitted offline and output downloaded later.
A VS Code extension that turns a remote GPU server into what feels like a local one. Users upload their Python project, browse files, launch execution, and watch output stream directly into their editor — all without leaving VS Code. Finished jobs can be downloaded in a single click.
A customized VS Code distribution with multi-workspace tabs and integrated project configuration, build, and debugging for C/C++ and Python projects. It ships with the MTCode RemoteGPU and AI Phoenix extensions pre-installed.
A self-hosted AI coding assistant for VS Code and MTCode Studio — inline completion and chat powered by your own LLM server, with nothing sent to third-party clouds. It has a built-in core library shared with MTCode Portal and can operate independently without it. Ideal for small software projects and for beginners and students asking programming questions.