Stop debugging the same broken agent loop twice.
A curated pattern library for Claude agentic workflows. Searchable cards with code, failure modes, and flow diagrams.
Pattern cards
Problem, flow diagram, copy-paste snippet, and failure modes all in one card.
Instant search
FuseJS runs in the browser with no backend. Find the pattern you need in under a minute.
Documented failure modes
Every pattern names the anti-patterns and edge cases that burn tokens or break loops in production.
How it works
Search for your pattern
Type a keyword and FuseJS returns matching cards instantly, no backend required.
Read the card
Each card shows the problem, a flow diagram, a copy-paste snippet, and known failure modes.
Copy and ship
Paste the snippet into your project and skip the debugging cycle that slows everyone else down.
Before & after
- Reinventing agent loops that break in production
- Hunting Reddit threads for unvetted, unsearchable advice
- Burning tokens on anti-patterns you find only after they break
- Proven patterns with failure modes documented before you ship
- One searchable source you can trust instead of ten blog posts
- Know every failure mode before you write a line of code
FAQ
Is this specific to Claude Code?
The patterns target Claude Code and Claude API agentic workflows. Most also apply to any LLM pipeline using tool calls, context management, or multi-step reasoning.
How is this different from the official Anthropic docs?
Anthropic docs cover API reference. This covers production patterns: what breaks, how to structure loops, how to recover from errors, and how to manage context at scale.
Can I use this without an account or API key?
Yes. The site is fully static. Search runs client-side with FuseJS. No account, no API key, no tracking. Just open it and search.