Fadé Adeniyi
Thursday 17 September 2026
If you're using Claude Code, install these five. One scans your codebase and tells you what to set up. One gives you hundreds of free model fallbacks when you hit a limit. One finds the right skill for whatever you're building. One attacks your own app like a real hacker to find the holes first. One reads the internet for you without a single API key.
Out of the box, Claude Code is a blank terminal with a very capable model behind it. Everything that makes it feel like a real setup, the fallback options, the security checks, the research abilities, comes from what you choose to add. Most people either add nothing, because they don't know what's worth adding, or add too much, because a random list told them to.
Below are five plugins, what each does, and the real command to install it. All are free.
If you already run Claude Code plugins, skip ahead.
It reads your whole codebase and recommends the MCP servers, skills, hooks and subagents that fit your project, then tells you what to skip. It runs read-only, so it never changes a file on its own.
/plugin marketplace add anthropics/claude-plugins-official /plugin install claude-code-setup@claude-plugins-official
Claude Code usually adds Anthropic's own marketplace automatically the first time you run it, so the first line may already be done. Once installed, ask: "recommend automations for this project."
OmniRoute sits between Claude Code and the model providers. Connect it to a handful of free providers and it switches over the moment your usual model runs out or gets rate limited, so you keep working instead of waiting.
npm install -g omniroute omniroute setup-claude-code
The second command configures Claude Code for you against OmniRoute's local gateway. Connect at least one free provider from its dashboard at localhost:20128 before you rely on the fallback.
Built by Vercel, this works like a package manager for skills. Describe what you're building and it searches known skill marketplaces for the ones that would help, instead of you scrolling a list guessing.
npx skills add https://github.com/vercel-labs/skills --skill find-skills
Once it's in, ask something like "what skills are useful for this project?" and it does the searching.
Strix runs your app and tries to break it, producing working proof-of-concept exploits instead of a list of theoretical warnings. It's built for the apps people vibe-code fast and ship without a proper security pass.
curl -sSL https://strix.ai/install | bash export STRIX_LLM="your-model-here" export LLM_API_KEY="your-api-key" strix --target ./your-app-directory
Docker needs to be running. The first scan pulls a sandbox image, so it takes a minute longer than the ones after it.
Agent Reach gives Claude access to web pages, X, Reddit, YouTube, GitHub and more, without you paying for or managing a single API key.
Install Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md
Paste that line straight into a Claude Code conversation. Claude reads the install doc and sets itself up, checking your machine first rather than writing config without asking.
None of the five overlap. Each covers a gap the others don't.
Each plugin shows up when you run /plugin in Claude Code. Ask Claude Code Setup for recommendations and it should reply with real suggestions for your codebase, not a generic list.
The trap is installing all five at once because they're free. The best setup isn't the one with the most plugins. It's the one where each plugin does a job the others can't.
Being technically able to scrape a platform isn't the same as being allowed to under its terms of service. Check what you're permitted to pull from a site before you point Agent Reach at it. Star counts, provider counts and commands change, so check each project's own page for current details.
Sources and honesty notes: Claude Code Setup is confirmed in Anthropic's own claude-plugins-official repository. OmniRoute, Strix, find-skills and Agent Reach are each real, independently published open-source projects, verified against their own repositories and install docs at the time of writing. Star counts and provider counts for these tools change, so check each project's own page for current numbers. Commands shown are current as of this write-up and may change between versions.
The Audacity to Build is my community for people doing exactly that: building real AI systems alongside me, step by step. Join the waitlist and I'll tell you the moment doors open.
If this was useful, follow me. I share what I am actually building, the tools that survive my own testing, and the ones I quietly cancel. New plays every week.
Pop in your name and email and the rest of the guide is yours, along with the occasional new guide when I publish one. No spam, ever.