netrun-mcp
This server allows you to deploy, manage, and monitor code projects hosted on the Netrun platform.
Deploy local project folders (e.g., Node, Python, Docker) to Netrun, automatically building them and providing a public URL.
Manage project lifecycle by starting, stopping, or restarting applications.
Configure encrypted secrets (API keys/tokens) for your project and trigger a re-publish to apply changes.
View real-time logs, project status, and manage custom domain settings.
Monitor your Netrun account usage, including plan details and remaining free running time.
Allows publishing Telegram bots to Netrun, including setting bot tokens as secrets and managing new versions of the bot deployment.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@netrun-mcp publish the current folder to Netrun"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
netrun-mcp
Publish websites and Telegram bots to Netrun — upload a folder from your editor, get a URL.
Connect Netrun to your AI editor and just ask: “publish this bot”. The agent uploads your code folder and sends back the link. MCP is the common way to connect external services to AI editors. With Netrun connected, you tell the agent in chat that you want the project online: it uploads the code folder, asks for a token if the code needs one, waits for the build and gives you the link. Works in Cursor, Claude Code, Codex, Claude Desktop, VS Code and Windsurf, and in ChatGPT and Claude right in the browser.
Publish this bot to Netrun
I fixed the greeting — update the project
Why is the bot silent? Check the logs
Change the bot token and restart
Setup (one minute)
Sign in at netrun.io → MCP (https://netrun.io/mcp) → create an access key.
Connect your editor. The MCP page gives you a ready button or command with the key already inside; the generic form is:
{
"mcpServers": {
"netrun": {
"command": "npx",
"args": ["-y", "netrun-mcp@latest"],
"env": {
"NETRUN_API_KEY": "nru_…"
}
}
}
}Cursor — click Add to Cursor on the MCP page, or put the JSON above into
~/.cursor/mcp.json.Claude Code —
claude mcp add netrun -s user -e NETRUN_API_KEY=nru_… -- npx -y netrun-mcp@latestCodex —
codex mcp add netrun --env NETRUN_API_KEY=nru_… -- npx -y netrun-mcp@latestClaude Desktop, VS Code, Windsurf — the JSON above in the MCP servers config.
Node.js has to be installed — AI editors usually ship with it.
ChatGPT, Claude in the browser — no install, no key
Netrun also runs as a remote MCP server: https://api.netrun.io/v1/mcp (Streamable HTTP, OAuth sign-in). In the chat settings open connections (Connectors or Apps), add a custom one with this address and press Connect. You sign in to Netrun by email — no key needed. ChatGPT allows custom connections on paid plans — turn on developer mode in the settings. Claude allows them on every plan, one on the free plan. The agent takes the code right from the conversation; a large multi-file project is easier to publish from an editor.
Related MCP server: localmcpcoder
What the agent can do
Tool | What it does |
| Uploads a project folder and puts it online — a new project or a new version of an existing one. Detects the stack (Python, Node, Go, Rust, PHP, static HTML, Docker, docker-compose), asks for missing secrets, waits for the build and returns the URL or the failure reason. |
| Is the project running, its URL, what blocks it, which secrets are already set and why the last publish failed. |
| Last lines of the application output — the first place to look when a bot goes silent. |
| Saves tokens and API keys as environment variables (stored encrypted) and re-publishes so the app picks them up. |
| Decides whether the project gets a public HTTPS address — for example a webhook bot we took for a background program. Re-publishes the project; the link appears if the app answers requests from the internet. |
| Points your own domain at the site and checks whether the registrar settings have taken effect. Until they have, it shows which records to add. |
| Start, stop or restart a project. |
| All projects of the account with URLs and state. |
| The plan, how many projects are available and how much free running time bots have left. |
What it deliberately cannot do
The agent cannot pay, renew a plan or buy places — you do that in the dashboard.
It cannot delete projects or files from persistent storage either — anything irreversible stays with you.
If the code needs a token or a password, the agent asks you before publishing — it never makes values up.
The access key lives only in your editor config, and a connected chat is listed in the dashboard — you can revoke either there at any time.
How it works
This package is a thin courier: it zips the folder (honouring .gitignore), sends it to the Netrun API with your key, and asks for status and logs. Building, running, HTTPS and addresses all happen on Netrun.
Environment variables: NETRUN_API_KEY (required), NETRUN_API_URL (default https://api.netrun.io/v1), NETRUN_APP_URL (default https://netrun.io).
Registry:
io.netrun/netrun
MIT
Available Tools
7 toolsnetrun_controlStart / stop / restartBInspect
Start, stop or restart a project. Stopping a Free-plan bot does not pause its free running time.
Inputs: action (required), project_id (required).
netrun_list_projectsList Netrun projectsARead-onlyInspect
All projects of the account with id, kind, status and public URL. Use it to find project_id for other tools.
Inputs: no named parameters.
netrun_logsProject logsARead-onlyInspect
Last lines of the application output (stdout/stderr) — the first place to look when a bot is silent or a site errors.
Inputs: project_id (required), lines.
netrun_publishPublish to NetrunAInspect
Upload a project folder and put it online: creates a new project (pass name) or ships a new version of an existing one (pass project_id). Detects the language automatically (Python, Node, Go, Rust, PHP, static HTML, Docker, docker-compose…). Waits for the build and returns the public URL or the failure reason. If it returns needs_secrets, ask the user for the values and call again with "secrets".
Inputs: path (required), name, wait, comment, +4 more.
netrun_set_secretsSet project secretsAInspect
Save environment variables (tokens, API keys, passwords) for a project. Values are stored encrypted and never returned. By default the project is re-published so the app picks them up. Ask the user for the values — do not invent them.
Inputs: secrets (required), project_id (required), file, publish.
netrun_statusProject statusARead-onlyInspect
Current state of one project: is it running, its URL, what blocks it, and — if the last publish failed — why (stage, reason, stderr tail).
Inputs: project_id (required).
netrun_whoamiNetrun accountARead-onlyInspect
Who the configured key belongs to, the current plan, project limit and — on the Free plan — how much free running time bots have left. Call it first when the user asks anything about their Netrun account or limits.
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 7 tools
Each tool targets a distinct aspect: account info, project listing, per-project status, logs, publishing, secrets, and control actions. There is no overlap in purpose, and descriptions clarify when to use each.
All tool names use snake_case with a clear 'netrun_' prefix, but the pattern is mixed: some are verbs (publish, control), some nouns (status, logs), and 'whoami' is a command. Still, names are intuitive and predictable, with minor deviation from a strict verb_noun convention.
Seven tools is well-scoped for a deployment platform sever. Each tool covers a necessary function without redundancy, and the count is within the ideal range.
The tool set covers the core lifecycle: account info, project discovery, status, logs, publish/update, secrets, and start/stop/restart. A delete project operation is absent, but this is a minor gap that does not break typical workflows.
Maintenance
Related MCP Connectors
Build, deploy, and host full-stack web apps from any MCP client. DB, auth, storage, cron included.
Publish websites with a small backend (KV, SQLite, files). MCP for AI agents.
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
Deploy AI-generated HTML/CSS/JS to instant public HTTPS URLs from any MCP-compatible agent.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceMCP server for self-hosted static site publishing on Cloudflare Workers. Enables AI coding agents to deploy pages with a single 'publish' tool and get live URLs, with support for atomic updates, versioning, and per-site passwords.8 npm2MIT
- AlicenseNot gradedqualityDmaintenanceEnables remote MCP clients like ChatGPT to run shell commands and manage files on your local machine via a Cloudflare tunnel, exposing tools for file operations, search, and task management.7MIT
- AlicenseNot gradedqualityDmaintenanceDeploy GitHub repositories or local directories to private URLs directly from an MCP client. Includes tools for managing deployments, reading build logs, and stopping services.43 npmMIT
- AlicenseNot gradedqualityBmaintenanceEnables MCP-capable assistants to build, preview, and publish full-stack web apps by directing a hosted coding agent, with tools for projects, wallet balance, domains, previews, downloads, and GitHub export.MIT