16,000+ MCPs, plugins, and skills exist for Claude Code — and growing every week. Claude Code only sees what's installed. XFTD scans the full catalog, ranks every tool against your actual task and stack, and surfaces the best one — whether you have it installed or not.
Connect with GitHub — it's freeEvery time your prompt changes direction — from debugging to UI to deployment to database work — XFTD classifies the shift automatically. It doesn't wait for you to declare what you're doing. It knows within one message.
Skills, plugins, MCPs — all indexed and pre-scored. XFTD detects your project's languages, frameworks, and installed MCPs, then ranks tools against that profile — not just by popularity. A Flutter project gets Flutter-specific rankings. A Postgres project gets database-specific rankings.
The top tools appear where you're already working, grouped by type (Plugin / Skill / MCP), with scores and install commands. You see what's available before Claude Code chooses. You decide whether to use one.
Before every tool call, XFTD scores the tool Claude Code selected against the best available for your task. If the gap is 10+ points, XFTD blocks the call and surfaces a better option. Not a nudge — a hard stop. You can override, but Claude Code can't ignore it.
BYOK and Free tier use live search against skills.sh and glama.ai — accurate and current. Pro uses a pre-ranked catalog updated daily from 6 sources, delivering results in under 200ms — with better LLM-driven ranking for complex stacks.
You submit a prompt — UserPromptSubmit hook fires.
Task type is classified in under 1 second against 16 canonical categories.
Evaluator searches the catalog and ranks tools against your detected stack profile.
Recommendations appear in the conversation, grouped and scored.
When Claude reaches for a tool — PreToolUse fires. XFTD scores it against the best available.
10+ point gap: Claude's tool call is blocked. Better option surfaced. You decide.