Your AI coding assistants could be so much better.

You know it. You've seen glimpses of it. But getting there means fighting config formats, duplicating setup across tools, and constantly babysitting permissions.

agentrc fixes that.

The problem isn't the AI. It's the setup.

You've got Claude Code, Cursor, maybe Windsurf or Copilot. Each one has its own config format. Its own quirks. Its own way of defining what the assistant should and shouldn't do.

So you end up with:

agentrc lets you teach your AI assistants how you work. Define your preferences once — in plain language or structured config, your choice — and agentrc handles the rest. It translates your intent into native formats, syncs across every tool you use, and gives your assistants intelligent boundaries so you stay in flow instead of approving every command.

What changes

permissions
Without agentrc
$ rm -rf node_modules
⚠ Allow? [y/n]

$ npm install
⚠ Allow? [y/n]

$ npm run build
⚠ Allow? [y/n]

$ npm test
⚠ Allow? [y/n]
With agentrc
$ rm -rf node_modules
✔ approved (regenerable)

$ npm install
↻ redirected → pnpm install
✔ approved

$ rm -rf node_modules/../../..
✖ blocked path traversal

agentrc understands command semantics. Safe operations flow through. Dangerous ones don't.

config
Without agentrc
# .cursorrules
"Always use pnpm for package management"

# CLAUDE.md
"Use npm for all package operations"

# .windsurfrules
(forgot to set this up)
With agentrc
# .agentrc
package_manager: pnpm

→ Applied to: Claude Code, Cursor, Windsurf
→ Last synced: 2 minutes ago
→ 0 drift detected

Define once. Sync everywhere. No more copy-paste. No more drift.

teaching
Without agentrc
# 47 lines of YAML copied from a blog post
rules:
  - pattern: "src/api/**"
    actions:
      - type: validate
        schema: input
        # ... 40 more lines you don't fully understand
With agentrc
"When working in /src/api, always validate input schemas
before processing. Never modify files in /config/prod
without asking."

→ 2 rules generated
→ Applied to all assistants

Describe what you want in plain language. agentrc writes the config.

Is agentrc for you?

agentrc is for you if:

  • You're already using AI coding assistants regularly — and believe they could be much more capable
  • You've felt the friction of configuring multiple tools, or wished your setup would just sync
  • You want to teach your assistants how you work without becoming an expert in five config formats
  • You'd rather spend time coding than babysitting permissions

agentrc is probably not for you if:

  • You're just getting started with AI coding assistants
  • You only use one tool and its built-in config works fine
  • You enjoy hand-tuning YAML files (no judgment)

How it works

agentrc is three capabilities working together:

SyncYour configuration, translated for every tool. Define preferences in .agentrc and Sync generates native config for Claude Code, Cursor, Windsurf, and more. Changes propagate automatically.
ShapeNatural language to structured config. Describe what you want in plain English. Shape converts it to precise rules your assistants can follow.
FlowIntelligent permission boundaries. Instead of approving every command, define policies once. Flow evaluates commands semantically and acts within the lines you've drawn.

Ready to try it?