Boolsai — MCP server suite

Stack-intelligence MCPs for AI agents. Live, no-auth, public access to every commerce site's tech stack — every vendor, every account ID, every inline script signature, plus a quant-research signal layer on top.

Built for AI agents doing competitive intel, agency prospect research, ecommerce due diligence, and ad-hoc commerce-data discovery.

This repo contains the Anthropic MCP Registry manifests for the four-server Boolsai suite. The actual servers live at *.boolsai.ai.

The four servers

Each server is independently published to the Anthropic MCP Registry under the ai.boolsai/* namespace (DNS-verified via boolsai.ai).

Why Boolsai

• No auth, no key, no rate limit gymnastics. Paste a URL into Claude.ai → Settings → Connectors. Done.
• First-party data. Every scan runs live on demand. No third-party feeds, no stale catalogue.
• Built for AI consumption. Each server's initialize response includes a detailed instructions block — vendor recognition tips, output schemas, methodology. Agents don't have to guess.
• Suite, not silo. Cross-discovery: each server's instructions reference the other three so agents can chain queries (scan → look up in directory → grep historical → test as signal).

Install

Claude Desktop / Claude Code / Cursor

{
  "mcpServers": {
    "boolsai-scan":      { "url": "https://boolsai.ai/mcp" },
    "boolsai-directory": { "url": "https://directory.boolsai.ai/mcp" },
    "boolsai-grep":      { "url": "https://grep.boolsai.ai/mcp" },
    "boolsai-signals":   { "url": "https://signals.boolsai.ai/mcp" }
  }
}

Restart the client. All four servers show up with their tools.

ChatGPT / Claude.ai Custom Connectors

Settings → Connectors → Add custom connector → paste any of the four URLs above. Add all four for the full suite.

One-liner test from a terminal

curl -sS https://boolsai.ai/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | jq .

Example agent prompts

"What's the full tech stack of gymshark.com?" → boolsai_scan({ url: "https://gymshark.com" })

"List ten stores using Klaviyo with their company IDs." → boolsai-directory.sites_using_vendor({ vendor: "klaviyo" })

"Show me sites with window._fs_org (FullStory) in their inline scripts." → boolsai-grep.sites_with_signal({ pattern: "window._fs_org" })

"Find the highest-α tradeable signal in the last 30 days with n ≥ 20." → boolsai-signals.find_signals({ horizon_days: 30, min_n: 20 })

Architecture

                   ┌──────────────────────────┐
                   │  AI agent (Claude / etc) │
                   └────────────┬─────────────┘
                                │ JSON-RPC 2.0 (Streamable HTTP)
            ┌───────────────────┼─────────────────────┐
            ▼                   ▼                     ▼
   boolsai.ai/mcp     directory.boolsai.ai/mcp   grep.boolsai.ai/mcp
   (live scanner)     (indexed corpus query)     (raw R2 search)
            │                   │                     │
            │                   │                     │
            ▼                   ▼                     ▼
     ┌──────────────────────────────────────────────────────┐
     │  R2 corpus (hundreds of thousands of historical scans) │
     └──────────────────────────────────────────────────────┘
                                ▲
                                │ derived
                signals.boolsai.ai/mcp  (quant signals from stack changes)

All four servers run on Cloudflare Workers with R2 for the historical corpus.

Publishing to the MCP Registry

See PUBLISHING.md for the one-time DNS verification + per-server publish commands.

License

MIT — see LICENSE.

Built by Boolsai · Contact: founder@boolsai.ai