Floom Library
📁
pr-review
v1.2
📁
bug-workflow
v2.0
+
9 more
synced
OPEN SOURCE · MIT
One skill library,
synced across
every agent.
Same library, every agent. Syncs instantly.
Claude Code
✓ synced
Codex CLI
✓ synced
Cursor
✓ synced
Kimi
✓ synced
Gemini
✓ synced
OpenCode
✓ synced
One library per project. Folders inside. Skills inside folders. Toggle libraries on or off — your agent reads only what's active.
WORKS WITH
Claude
Codex
Cursor
Kimi
Gemini
OpenCode
TRUSTED BY
"Saved my team from shipping a broken migration last week."
AV
"We replaced our internal review checklist with this."
TK
"Pattern-sweep alone is worth it."
PS
Save once.
Publish a skill folder from any machine. The Floom MCP picks it up — a single Node binary, boots in 200ms, idles at 30MB. Lightweight by design.
Sync everywhere.
Use Cursor today, Claude Code tomorrow, Codex on a different machine next week. Same library, no re-setup. Zero switching cost across agents.
Search live.
Skills don't need to be pre-loaded. The MCP exposes search_skills and get_skill as tools — agents discover and load mid-task.
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 + power users (init / publish / sync). Receivers can install via the web (visit a /s/ link, click "Save to my floom") — no terminal required.
Does it work with my agent?
Native: Claude Code, Codex CLI. Via MCP: Cursor, Kimi, Gemini, OpenCode. New agents added as MCP support lands.
Where does my data live?
Your library lives on Floom (Supabase-backed, source of truth). The actual skill files live on each machine you sync to. Your data is yours — export anytime.
Is this open source?
Yes. MIT-licensed, source on GitHub at floomhq/floom. Contributions welcome.
Save your first skill.
Sign in, browse the library, save — done.