While Claude and Cursor get all the headlines, Google’s Gemini models (especially the 1.5 Pro and Flash series) are incredible at handling massive contexts. And thanks to the community, you can use Gemini in the terminal with MCP support.
Why Gemini + MCP?
Gemini 1.5 Pro has a 2 Million Token Context Window. This is huge. When you combine this with MCP tools that can fetch massive amounts of data (like pulling entire documentation sites or large log files), Gemini shines. It can ingest a volume of data that would choke other models.
Setting Up the Gemini CLI
(Note: There are several CLI wrappers for Gemini. We will focus on the official Google Cloud SDK patterns or popular open-source wrappers that support MCP).
Assuming you are using a standard CLI wrapper that supports a config file (often YAML or JSON).
Configuration
The config usually looks similar to Claude’s but often prefers YAML.
tools:
- name: filesystem
command: npx
args:
- -y
- @modelcontextprotocol/server-filesystem
- /path/to/allowed/dirThe “Terminal Agent” Workflow
Using MCP in the CLI feels different. It’s like having a super-powered Bash shell.
Example Session:
$ gemini chat
> "Check the git logs for the last 24 hours, identify any commits that touched the 'auth' module, and list the authors."- Gemini calls the
gitMCP server. - It gets the raw log.
- It filters and summarizes the output right there in your terminal.
Syncing with Vibe Manager
Vibe Manager treats the Gemini CLI as just another “Target.” You can define your tools once (FileSystem, Git, Postgres) and sync them to your Gemini CLI config.
Vibe Manager handles the Format Conversion.
- It takes your Claude JSON configuration.
- It converts it to the YAML (or specific JSON structure) required by the Gemini CLI tool.
- It saves it to the correct path.
This allows you to experiment with Gemini 1.5 Pro in the terminal using the exact same toolset you use in Claude Desktop, without rewriting configuration files.