Core Agent Loop
From prompt to patch to verification.
Osmogrep runs as a local terminal agent with explicit tools. It inspects code, proposes edits, executes commands, and validates changes with tests before you ship.
Interactive TUI Session
Stream agent responses and tool logs in one place with clear separation between input, execution, and output.
Tool-Driven Editing
Use explicit tools like `read_file`, `search`, and `write_file` so every edit path stays auditable and reproducible.
Command + Control
Run shell directly with ! and manage execution with slash commands like /help and /quit.
index://graph-build
First Launch Indexing
Osmogrep builds repository context before any edit loop starts.
On first run, Osmogrep scans your workspace and builds a relation graph for files, symbols, imports, and call paths. That context is cached and reused so patch generation and test execution stay grounded in real dependencies instead of guesswork.
symbolsimportscall paths.context.json
OSMOGREP
runtime monitor
$ osmogrep watch --project ./backend --fix
[engine] runtime hooks initialized
[watch] auth.ts changed
[analyze] runtime trace graph updated
_
Runtime Coverage
Works where you work
Built as a terminal-first runtime tool that runs cleanly across operating systems and editor workflows.
Terminal-first and cross-platform
Windows
macOS
Linux
VS Code
Neovim
Emacs
IntelliJ