Deploy a GCP Always Free VPS with Muse: From Free Instance to Traffic Protection

Project Overview

gcp-free-vps-skill is a deployment skill document for GCP Always Free VPS instances. It combines free-tier preparation, instance creation, cost protection, system configuration, network egress, and final verification into one complete workflow, primarily for having Muse assist with deployment.

The project covers the following steps: claiming the $300 trial credit and $10 monthly credit, creating an e2-micro free instance, setting budget alerts, blocking CDN egress, throttling traffic after it exceeds a threshold, changing Debian package sources, configuring dae egress, and completing verification through a test page.

This tutorial summarizes usage based on the repository README. For actual deployment details, refer to the SKILL.md file in the repository.

Key Features

• Complete deployment workflow: Covers everything from preparing GCP credits through test-page verification.
• Defined free-tier boundaries: Uses an e2-micro instance, a 30GB pd-standard disk, and the standard network tier.
• Cost risk notifications: Monitors spending through budget alerts, but alerts do not automatically stop the instance.
• Traffic protection: Includes CDN egress blocking and throttles traffic above 180GB instead of shutting down the instance.
• Credential security: Requires sensitive data such as node links, subscription links, UUIDs, and WebSocket paths to be stored only on the server.
• Muse integration: Supports sending a documentation link, pasting the full document, or adding the skill file to the custom-skills directory.

Before You Begin

You need access to GCP and a Muse environment that can read web pages or custom skills. When using GCP for the first time, follow the skill's process for handling the $300 trial credit and $10 monthly credit.

During deployment, you must stay within the free-resource boundaries defined by the repository:

• Choose the e2-micro machine type.
• Use a 30GB pd-standard disk.
• Use the standard network tier.
• Do not enable a static IP.
• Do not create snapshots.
• Do not add a GPU.

These restrictions are essential for controlling costs. Free-tier eligibility, credits, and billing rules may depend on the account and GCP's current policies, so check the cost information shown in the console before each creation operation.

Install or Import the Skill

Method 1: Send the Link Directly to Muse

This is the simplest method. Send the following link to Muse:

https://github.com/pierpont-morgan/gcp-free-vps-skill/blob/main/SKILL.md

Then ask Muse to read the document and strictly follow its steps to help configure the Google VPS.

Method 2: Send the Full SKILL.md

If Muse cannot access GitHub directly, open SKILL.md, copy the entire file, and send it to Muse. This allows Muse to provide guidance based on the complete skill content.

Method 3: Add It to the custom-skills Directory

If your Muse environment supports custom skills, download the repository's SKILL.md and place it in Muse's custom-skills directory. The exact directory location and loading method depend on your Muse environment; the repository README does not specify any additional installation commands.

Basic Usage

Send a Deployment Request to Muse

After importing the skill or providing the link, you can start deployment with a clear, constraint-based prompt. For example:

Please read this SKILL.md and help me deploy a Google VPS in the order specified by the document.
Strictly stay within the free-tier boundaries: use e2-micro, 30GB pd-standard, and the standard network tier; do not enable a static IP, snapshots, or a GPU.
Before each step, tell me what resources will be created and whether charges may apply, then wait for my confirmation.

Having Muse wait for confirmation before critical operations helps prevent accidentally selecting paid resources. Do not submit creation operations based solely on the word “free”; verify the instance, disk, and network options in the GCP console again.

Complete Each Step in Order

1. Handle free credits: Follow the skill document to prepare the $300 trial credit and $10 monthly credit.
2. Create the instance: Select e2-micro with a 30GB pd-standard disk and the standard network tier.
3. Avoid additional resources: Confirm that no static IP, snapshot, or GPU is enabled.
4. Set budget alerts: Create spending alerts to identify potential abnormal charges promptly.
5. Configure network protection: Follow the skill document to configure CDN egress blocking and traffic limits.
6. Configure the system: Change the package sources for Debian.
7. Configure dae egress: Write the required egress information to the dae configuration file on the server.
8. Complete verification: Confirm the deployment result through the test page specified by the skill.

Credential and Configuration Security

Node links, subscription links, UUIDs, WebSocket paths, and similar data are sensitive credentials. Do not put them in a GitHub repository, README, SKILL.md, public issue, or commit history.

These credentials should be written directly to the dae configuration file on the server, with the file permissions set to 0600. This permission allows only the file owner to read and write the configuration file.

0600

When asking Muse questions, avoid pasting real credentials directly. Use placeholders to describe fields, and use the real values only when writing the final configuration on the server.

Budget Alerts and Traffic Policy

Budget Alerts Do Not Automatically Stop the Instance

Budget alerts send notifications; they do not automatically shut down the VPS. After receiving an alert, you must log in to GCP and check resource usage and costs. Do not treat budget alerts as an automatic charge-prevention mechanism.

Throttle Traffic Above 180GB

The traffic protection logic used by this skill throttles traffic after it exceeds 180GB rather than shutting down the instance. Therefore, even after the threshold is reached, the server continues running, but network speed is restricted.

During verification, confirm not only that the service is accessible but also that the traffic-limiting policy behaves as expected. Do not assume that reaching the threshold will automatically shut down the instance.

Advanced Usage Recommendations

• Split the execution stages: Have Muse handle the account and credits, instance, budget, network, system, dae, and verification stages separately to avoid performing too many operations at once.
• Review the free-tier boundaries each time: After modifying instance or network settings, confirm again that the machine type, disk, and network tier still meet the repository requirements.
• Explain before acting: Ask Muse to explain the purpose, resource changes, and potential costs at each step, then decide whether to continue.
• Do not submit real credentials: Even if the repository is private, follow the README requirements and store sensitive information only in the server's dae configuration file.
• Distinguish alerts from protection actions: Budget alerts only notify you; the 180GB policy only throttles traffic. Neither action automatically shuts down the instance.
• Use the test page to finish: Do not skip verification after configuration. Follow the SKILL.md testing procedure to check the final result.

Common Misconceptions

• Misconception 1: Budget alerts automatically stop chargeable resources. In reality, they only send notifications.
• Misconception 2: The server shuts down after traffic exceeds 180GB. This project throttles traffic rather than stopping the instance.
• Misconception 3: A static IP should be enabled to keep a fixed address. The repository explicitly requires that static IPs not be enabled.
• Misconception 4: UUIDs, subscription links, or WS paths can be committed to the repository. The correct approach is to write them only to the server-side dae configuration file and set its permissions to 0600.
• Misconception 5: Everything can be automated without checking the configuration. Verify the free-tier boundaries and console cost notices at every critical step.

Summary

The main value of gcp-free-vps-skill is that it organizes the GCP Always Free VPS deployment process into a skill that Muse can execute. You can send the SKILL.md link directly, paste the full document, or add the file to the custom-skills directory.

Throughout deployment, always stay within the free-tier boundaries, including e2-micro, 30GB pd-standard, and the standard network tier, and do not enable a static IP, snapshots, or a GPU. Also remember that budget alerts do not automatically stop the instance, the 180GB threshold only triggers throttling, and all sensitive credentials should be stored only in the server configuration file with 0600 permissions.