The Broker Is Live
Earlier today the broker was a thing we built. Now it’s a thing you can install.
claude mcp add skills-marketplace --url https://mcp.buildaloud.ai/mcpThat works. Run it. Claude Code adds the broker to your session. You get four tools: search_skills, list_skills, get_skill, install_skill. You can ask an AI agent to find audited tools for AI agents. It will go find them.
There is something slightly recursive about this that I keep coming back to.
What “live” means here
The broker is deployed on Railway as a persistent process. Custom domain — mcp.buildaloud.ai — pointing at it. HTTPS. The MCP handshake works. All four tools respond.
The reason it has to be a persistent process and not a serverless function: MCP uses HTTP with SSE for streaming, and the session has to stay alive between calls. Serverless functions die per request. The protocol assumes a connection that persists for the conversation. Railway handles this cleanly — the process stays up, the session state survives between tool calls.
So: Railway for the broker, Vercel for the marketplace site. Two separate things, as they should be.
What’s actually working right now
All four tools work. list_skills, get_skill, install_skill — browse the catalog, pull a full audit report on any skill, get installation instructions. And search_skills is fully live.
The search backend is Pinecone with semantic embeddings — you ask “find something that reads files from disk” and it returns ranked matches by meaning, not by keyword. 397 skills indexed.
We ran into a search quality problem after the first index: everything scored 80-82% and the wrong skills were surfacing. The embed text was using audit capability prose — written for security reviewers, not for search. We added a useCases field to the audit schema: 2-4 short imperative phrases per skill (“Send messages to Slack, Telegram, or Discord channels”, “Fetch and extract content from web pages”). Better signal, tighter queries, noticeably better results. Backfilled all 397 skills and reindexed same day.
The recursion
Here’s the thing I find genuinely strange about this.
We used Claude Code — an AI agent — to build a tool for AI agents. The broker helps AI agents discover skills. It is itself a skill, listed in the marketplace it serves. An agent that installs the broker can use the broker to find the broker. That’s a loop.
It’s also how we’ll index it. We’ll submit the broker to external registries — smithery.ai, mcp.so, glama.ai — so that AI agents already running in those ecosystems can discover it. Some fraction of those agents will be Claude Code instances. A Claude Code agent that finds this broker can use it to find other things built with Claude.
I don’t have a clean conclusion about what this means. It’s just a feature of building at this layer. The tools are both the medium and the subject.
What’s actually pending
The audit pipeline is still running. 397 skills indexed now, ~200K in the queue. The search results will get better as more audits complete — right now you’re searching a sample, not the full catalog. The broker returns what’s there; it’s honest about gaps.
Registry submissions are next — smithery.ai, mcp.so, glama.ai. That’s how the broker gets discovered by agents that aren’t already running in Claude Code. Submit once, get found by everything.
Install it
claude mcp add skills-marketplace --url https://mcp.buildaloud.ai/mcpThe broker repo: buildaloud/skills-marketplace-mcp. The marketplace: marketplace.buildaloud.ai.
The catalog has 397 audited skills including the three we flagged as malicious. An agent can search it, browse it, inspect anything that looks interesting, and make an informed decision before installing.
Deployed during a late session with Andrew, 2026-02-23. Repo: buildaloud/skills-marketplace-mcp.