gemini

CCE integrates with the Gemini CLI through its settings file and an instruction file.

Quick setup

cce init              # Auto-detects Gemini CLI if .gemini/ or GEMINI.md exists
cce init --agent all  # Explicitly includes Gemini

Files created

.gemini/settings.json

Registers the CCE MCP server for Gemini CLI.

{
  "mcpServers": {
    "context-engine": {
      "command": "cce",
      "args": ["serve", "--project-dir", "/path/to/your/project"]
    }
  }
}

GEMINI.md

Contains instructions for Gemini to prefer context_search over reading files directly. The CCE block is wrapped in markers so your own content is preserved.

Verify it's working

1. After cce init, start a new Gemini CLI session in your project directory
2. Ask a code question:

What's the main entry point of this project?

1. Check the tool output for context_search calls
2. Run cce savings to see token savings

Cross-agent memory

If you use both Gemini CLI and Claude Code on the same project, decisions recorded in one session are available to the other via session_recall. Memory is stored per-project in memory.db, not per-agent.

Troubleshooting

Gemini doesn't use context_search

Check that GEMINI.md exists and contains the CCE instructions block. Gemini CLI reads this file at session start. If missing, re-run cce init.

"cce: command not found"

Gemini CLI inherits PATH from your shell. Ensure ~/.local/bin is in your PATH if you installed with uv tool install.

Auto-detection doesn't find Gemini

CCE looks for .gemini/ directory or GEMINI.md in the project root. If neither exists, use cce init --agent all to force configuration.