Grep files

grep_files searches the text of every file below a folder, returning matching lines with file and line number.

Usage

For "where did I write that?" questions, searching beats reading one file at a time:

Search ~/notes for everywhere the invoice is mentioned

A ~ is expanded, and relative paths resolve against the directory Mycel was started from — so a full path is always the clearer thing to give. Follow up with read_file once a match points you at the right one.

Scope and limits

It only reads. Nothing here writes, moves or deletes a file.

• A search skips machine-generated folders — .git, node_modules, vendor, .venv, __pycache__ — and stops after 50 matching lines or 2,000 files, so pointing it at something enormous degrades instead of hanging.
• Long lines are shortened in search results, to keep a match readable rather than dumping a minified file into the reply.

Any path your account can read

This tool is not confined to a folder: whatever your user account can read, Mycel can search if it is asked to, including ~/.ssh, ~/.aws and ~/.config/mycel/.env. That matters because Mycel takes instructions from anywhere it is reachable — a Telegram bot is public unless you keep its token private, and a web page Mycel reads can carry text that tries to talk to it. Keep the bot token to yourself, and think of the file tools as being as trusted as the people who can message the agent.

The model needs tool support

Tool calling depends on the model. If LLM_MODEL has no tools capability, Mycel warns at startup and grep_files is ignored along with the rest.