I experimented to see if I could create a similar environment without the Claude Code subscription cost. By attaching MCP servers to Gemini CLI, you can create a premium-level coding workflow for free. This guide covers everything from installation to actual usage and troubleshooting.


First, Let's Clarify the Difficult Terms in 1 Minute

  • Gemini CLI: A command-line tool that lets you use Gemini models in the terminal. Enables conversations, code generation, and file context utilization.
  • MCP (Model Context Protocol): A standard for connecting AI tools (servers) with models to extend functionality. It's like a "plugin."
  • Context7: An MCP server that instantly finds documentation/references and injects them as context. It fixes code based on the latest docs.
  • Sequential Thinking: An MCP server that "structures thinking steps." It breaks down requirements into step-by-step plans.
  • FastAPI: A Python backend web framework. Fast and type-friendly.
  • Tailwind CSS: A utility-class-based CSS framework. Great for quickly building beautiful UIs.
  • Alpine.js: A lightweight JS framework for simple state/interactions.
  • SQLite: A super simple local DB that's just one file. Perfect for development/prototyping.
  • GEMINI.md: A project context file that Gemini CLI reads. Good for putting requirements/design/rules.
  • .gemini/settings.json: Per-project Gemini CLI settings file. MCP server registration is done here too.

Why the Context7 + Sequential Thinking Combo?

  • Sequential Thinking first establishes the overall design and steps. (Architecture → Backend → Frontend → Integration/Testing)
  • Context7 fills in the gaps with latest documentation. It immediately fixes issues like framework changes and dependency conflicts.
  • Using both with Gemini 2.5 Pro (Gemini CLI's engine) creates a premium coding environment for free.

![image.png](https://storage.aickyway.com/api/files/post-images/2025/09/12/4cbd40cf-a34f-4082-acfe-7bdff2d655d9_image.webp)

Getting Started: Installation and Basic Folder Structure

Prerequisite: Assuming Gemini CLI is already installed. Project-level configuration is recommended.

Example Folder Structure

your-project/
├── .gemini/
│   └── settings.json
├── GEMINI.md          ← Create in project root
├── src/
├── package.json
└── other files

Initial Setup Commands

# Create .gemini folder
mkdir .gemini

# Create project context file (VSCode version, any editor works)
code GEMINI.md

# Open settings file
code .gemini/settings.json

settings.json (Minimal Configuration)

{
  "contextFileName": "GEMINI.md"
}

## MCP Server Installation/Registration

Register MCP servers in settings.json. (Project-level makes team sharing easy.)

{
  "contextFileName": "GEMINI.md",
  "mcpServers": {
    "Context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp"]
    },
    "sequential-thinking": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
    }
  }
}

Tip: Project settings (.gemini/settings.json) take precedence over user settings (~/.gemini/settings.json). Safely lock down project-specific context.


Verifying MCP Operation

Check if the servers are detected with these commands:

  • /mcp — View all configured MCP servers
  • /mcp desc — See descriptions of each MCP
  • /mcp list — List available MCP functions

Documentation links will open and guide you through initial server setup. Make sure to match GEMINI.md and project root location.


Planning the App: Sequential Thinking's True Value

Prompt example:

"Let's build a full-stack app Promptstore. CRUD (create/read/filter/save). Backend: FastAPI, Frontend: Tailwind + Alpine.js (or vanilla), DB: SQLite. Before starting, design the overall architecture and steps."

Sequential Thinking generates a step-by-step plan like this:

  1. Architecture Planning: DB schema, API endpoints, frontend components, file structure
  2. Backend Development: FastAPI setup, SQLite initialization, CRUD implementation, route definition
  3. Frontend Development: HTML + Tailwind, Alpine.js interactions, form validation, dynamic rendering
  4. Integration & Testing: API integration, error handling, basic tests, deployment preparation

Dependency Issues? Use Context7 to Inject 'Latest Docs' and Fix Immediately

When the app runs but crashes due to dependency conflicts, try requesting this:

"I have dependency issues causing crashes. Use Context7 to pull up-to-date docs and fix the code."

Then Context7 pulls FastAPI latest documentation/guides inline and fixes code and settings. (Version changes, import paths, config keys, etc.)


Issues and Real-World Tips

  • Occasionally there was prompt duplication or freezing during VS Code integration. Still investigating whether it's an editor extension/integration issue or CLI issue itself.
  • Prevent context leakage with project-level settings: Rules from other projects don't mix in.
  • Write architecture/coding rules/checklists in GEMINI.md, and the model follows them consistently.
  • Automate small things too: Let the model handle CRUD scaffolding, type hints, test stubs.

Conclusion: Free but Powerful

  • Sequential Thinking: The brain that turns complex requirements into step-by-step plans
  • Context7: Instantly fetches latest documentation to precisely patch bugs/changes
  • Gemini 2.5 Pro: The engine for generation and problem-solving

This combination is a powerful alternative to Claude Code and a great workflow. What MCP combinations are you using? Share in the comments below!