Small Business Intelligence

by Brick & Mortar — a free, standalone remote MCP server for tearing down small businesses.

It ships analytical methodology, not data. Every tool returns a rigorously structured framework — a research procedure, an output schema, a quality rubric, caveats — and the calling model executes the research itself, with its own web search. This server never calls an external API, a database, or any data provider on your behalf. Your own AI subscription pays for all the inference; our marginal cost per call is effectively zero.

Live endpoint: https://sbi-mcp.small-business-intelligence-mcp.workers.dev/mcp (no auth required — see Deploy for standing up your own copy).

The eight tools

Tool	What it does
business_teardown	Full structured teardown of one named business — digital presence, review signal, competitive position, pricing posture, visibility gaps, prioritized recommendations. The flagship tool.
competitor_landscape	Maps the local competitive set for a category + metro: true competitors vs. adjacent players, positioning matrix, saturation signals.
review_intelligence	Mines public reviews: complaint taxonomy, theme extraction, sentiment trajectory, differentiators customers actually cite, red flags for buyers.
local_visibility_audit	Local search presence audit: map-pack factors, listing consistency, category selection, site fundamentals — a scored checklist.
pricing_benchmark	Builds a defensible local pricing comparison, including how to normalize across service bundles and what to do when prices aren't published.
broker_diligence_prep	Pre-diligence framework for brokers/buyers: SDE framing, multiple ranges (caveated, verify-live), red-flag checklist, seller questions.
market_opportunity_scan	Gap analysis for a category × metro: underserved demand, oversaturation, whitespace — from public signals only.
compose_report	Assembles the outputs of prior tool calls into one polished, client-ready report, matched to the audience.

Why this architecture

See ARCHITECTURE.md for the full write-up. Short version: @modelcontextprotocol/server v2 (McpServer) + agents' stateless createMcpHandler, mounted as a plain Cloudflare Worker fetch handler at /mcp. No Durable Objects, no database, no external API calls — genuinely zero marginal cost per call.

Local development

npm install
npm run typecheck    # tsc --noEmit
npm run dev           # wrangler dev — serves http://localhost:8787

wrangler dev runs against Miniflare's local KV simulation, so the usage ledger works out of the box with no Cloudflare account needed for local dev.

Testing with MCP Inspector

With wrangler dev running in a separate terminal:

# list all 8 tools
npx @modelcontextprotocol/inspector --cli --server-url http://localhost:8787/mcp \
  --transport http --method tools/list --format json

# call one
npx @modelcontextprotocol/inspector --cli --server-url http://localhost:8787/mcp \
  --transport http --method tools/call --tool-name business_teardown \
  --tool-args-json '{"business_name":"Example Cafe","city_metro":"Saint Paul, MN"}' \
  --format json

Or drop --cli --format json for the interactive web UI (npm run inspector). Confirm all 8 tools list with readOnlyHint: true and an outputSchema, and that a tools/call against each returns structuredContent matching it — verified this way during the build (no isError, no protocol-level rejection, tools/list count = 8).

Manual smoke test (no Inspector needed)

curl -s -X POST http://localhost:8787/mcp \
  -H "content-type: application/json" \
  -H "accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"business_teardown","arguments":{"business_name":"Example Cafe","city_metro":"Saint Paul, MN"}}}'

Deploy

wrangler login                                   # or set CLOUDFLARE_API_TOKEN
wrangler kv namespace create USAGE_LEDGER          # once — copy the id into wrangler.jsonc
npm run deploy

Ships to the free *.workers.dev subdomain by default — no DNS work, no paid plan. A custom domain is a one-line routes addition in wrangler.jsonc later.

Usage stats

Every real tool call increments an aggregate, non-identifying counter (stats:{tool_name}:{date} in the USAGE_LEDGER KV namespace — see src/middleware/stats.ts, disclosed in /privacy and /docs). To see which tools actually get used:

export $(grep -v ^# .env | xargs)
python3 scripts/usage_stats.py            # last 30 days
python3 scripts/usage_stats.py --days 7    # last 7 days

Monetization (built, not active)

Every tool passes through a single middleware seam (src/middleware/context.ts). Today, POLICY_MODE=allow_all (the shipped default in wrangler.jsonc) never denies a call. Flipping it to metered activates a per-identity daily call limit (POLICY_METERED_DAILY_LIMIT), backed by a Workers KV usage ledger, with clean, plain-language denials (never a broken-looking protocol error) that include an upgrade link. No authentication is required at launch; the seam is architected so OAuth 2.1 + PKCE can be added later without restructuring — see src/oauth/stub.ts and ARCHITECTURE.md's "OAuth: stubbed, not mounted" section.

Repo structure

src/
├── index.ts          # Worker entry — routes / , /privacy, /mcp
├── server.ts          # createServer(): builds McpServer, registers all 8 tools
├── env.ts              # Env (Worker bindings/vars) type
├── middleware/          # identity resolution, KV usage ledger, policy, withPolicy seam
├── oauth/                # OAuth 2.1 discovery handlers — written, not mounted (see above)
├── tools/                  # one file per tool, all sharing the FrameworkPayload shape in types.ts
└── pages/                    # landing (/) and privacy (/privacy) page HTML

License

MIT — see LICENSE.

Privacy

See PRIVACY.md (or /privacy on the live deploy).