Content
# Context-Killer MCP
A little MCP server that keeps big code generation out of Claude's context window.
When Claude writes a large file, every line flows back through the conversation and
eats tokens. This server hands the job to Google's Gemini instead: Gemini writes the
file straight to disk, and Claude only gets back a one-line "done". The code never
lands in the chat.
It's a single tool, `delegate_to_gemini_and_save`. You give it a prompt and a path;
it optionally reads a few files for context, asks Gemini for the code, and saves it.
> Only use this for small taks or if you just want to try it out its not fully polished
> and claude doesnt see the results which is good for saving Tokens but it may lead to
> Errors since gemini api is not as good in coding and should only be used for small simple Tasks.
## The tool
`delegate_to_gemini_and_save(task_prompt, target_file_path, context_files=None)`
- `task_prompt` – what you want built, in plain words
- `target_file_path` – where to save it (missing folders are created)
- `context_files` – optional list of files for Gemini to read first
## Setup
You'll need Python 3.11+ and a free Gemini key from
<https://aistudio.google.com/app/apikey>.
```bash
pip install -r requirements.txt
cp .env.example .env # then paste your key into .env
```
## Use it in Claude Desktop
Add this to your `claude_desktop_config.json`:
- Windows: `%APPDATA%\Claude\claude_desktop_config.json`
- macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
```json
{
"mcpServers": {
"context-killer": {
"command": "python",
"args": ["C:\\path\\to\\context-killer-mcp\\server.py"],
"env": { "GEMINI_API_KEY": "your-key-here" }
}
}
}
```
Restart Claude, then just ask: *"Use context-killer to write … and save it to …"*.
## Good to know
- It writes (and overwrites) files on disk, so glance at what Claude's about to do
before approving.
- Your prompt and any context files get sent to Gemini — don't point it at secrets.
- `.env` is git-ignored. Never commit your real key.
## License
MIT — see [LICENSE](LICENSE).
Connection Info
You Might Also Like
ai-native-pm-os
The exhaustive guide to mastering Claude for Product Managers. Build your...
Train-in-Silence
The first Task-Aware MCP server and automated VRAM calculator for LLM...
stacklit
108,000 lines of code. 4,000 tokens of index. One command makes any repo...
AppClaw
AI-powered mobile automation agent — describe what you want in plain...
pdf-mcp
Production-ready MCP server for PDF processing with intelligent caching....
kotadb
Local-only code intelligence API for AI developer workflows (Bun +...