OPEN SOURCE · MIT

One skill library,
synced across
every agent.

Same library, every agent. Syncs instantly.

Save your first skill →Browse starter pack
Claude Code
✓ synced
Codex CLI
✓ synced
Cursor
✓ synced
Floom LibraryPersonal · cloud library
📁
brand-voice
v0.4
📄SKILL.md1.2 kb
📂voice-samples/3 files
📂templates/2 files
📄tone-guide.md0.4 kb
📁
pr-review
v1.2
📁
bug-workflow
v2.0
+
9 more
synced
Kimi
✓ synced
OpenCode
✓ synced

One library per project. Folders inside. Skills inside folders. Toggle libraries on or off, and your agent reads only what's active.

WORKS WITH
Claude
Codex
Cursor
Kimi
OpenCode
STARTER PACKS

Useful on day one.

Save a pack, then tailor the skills to your team.

Floom

Floom starters

Brand voice, PR review, and team onboarding.

  • brand-voice
  • pr-review
  • team-onboarding
Browse pack →

Save once.

Publish a skill folder from any machine. The Floom MCP picks it up: one Node binary, fast boot, low idle footprint.

Sync everywhere.

Use Cursor today, Claude Code tomorrow, Codex on another machine next week. Same library, no re-setup.

Search live.

Agents discover and load skills mid-task through search_skills and get_skill instead of preloading everything.

Quick answers.

What is a Floom skill?
A folder package: a SKILL.md plus optional examples, scripts, and references. Versioned, portable, native to every supported agent.
Why do I need a CLI for this?
The CLI is for publishers and power users. Receivers can save from a public /s/ link without touching a terminal.
Does it work with my agent?
Claude Code, Codex CLI, Cursor, Kimi, and OpenCode are the launch target set.
Where does my data live?
Your library lives on Floom as the source of truth. The actual skill files live on each machine you sync to, and your data can be exported.
Is this open source?
Yes. Floom is MIT-licensed and published on GitHub.

Save your first skill.

Two minutes from terminal to synced.

Get started →Browse starter pack