If you use a CLI-based LLM like Claude Code or Cursor to help with writing, you know the friction: edit files in one place, run the LLM in another, copy-paste between them.
Anvil bridges that gap. It is a browser-based writing workspace that reads and writes the same plain markdown files your LLM works with from the terminal.
What it does
Anvil gives you:
- Block-level editing with inline review notes
- Evidence sidebar (Ctrl+E) to search research files while drafting
- Guided form fields with word counters, field guides, and tips
- Status tracking: not_started to first_draft to human_written to ai_refined to final
- Resume editor with split-pane HTML + live preview + PDF export
- Export copy-paste-ready text for application portals
The workflow
You write first drafts in the browser. Your LLM reads and refines the same markdown files from the terminal. Status tracking keeps you honest about what is human-written vs AI-refined.
Stack
Flask 3.1 + Vanilla JS + SQLite + Docker. No build step. No npm.
The CLAUDE.md includes a full setup guide. Paste one prompt into your CLI and it interviews you, builds your evidence base, configures form fields, and writes your resume.
Get started
git clone https://github.com/herakles-dev/anvil.git
cd anvil
docker compose up -d
MIT licensed. Issues and PRs welcome.
GitHub: herakles-dev/anvil
Top comments (2)
Solid work.
Thanks, please follow on github :)