BUSINESS MANAGEMENT IN A BOX · MODEL CONTEXT PROTOCOL

Run the whole
business with
one agent.

Most AI talks about business. Evolved runs one — end to end: quotes that price themselves, safety, receipts, dispatch, invoicing, and on-chain settlement. It runs a real Alberta company today, and spins up for any trade in one call.

▶ Watch it run itself Make it your business
83MCP tools · 16 domains
2OKX rails · x402 + X Layer
2human gates · both money
41tests · live testnet probe
service checking… tools — X Layer testnet probing… paid calls — github.com/kr8tiv-ai/evolved
booting the books…
The 90-second film

See the whole thing run.

One agent takes a service business from a texted photo to a paid, on-chain invoice — every figure captured live from this endpoint. Then try it yourself below.

1080p · 90sreal captured payloadsphoto-quote → lifecycle → on-site JHA → workbook → x402 + on-chain
Operate
See it work · 60 seconds

Watch one agent run a company.

Every button below hits the real live service. Nothing is faked; the data is synthetic and restores itself hourly. Start with Judge Mode — one click runs the entire story, hands-free.

hands-free · one click

▶ Judge Mode — the whole business, autopilot

One click runs the entire story against the live service: the books of a real company → a photo turned into a priced quote → the autonomous lifecycle (holding at both human money gates) → any-business spin-up → the workbook spine → the raw x402 on-chain payment → tomorrow's digest. About 60 seconds. Touch nothing.

hands-free field ops

🎙️ Talk to the crew radio

Deterministic voice-intent parsing. Try: “used two bags of crushed glass on the Jasper Ave job” · “open the FLHA” · “next stop?” · “remind me to grab couplers today”.

seconds, not site visits

📸 Photo-to-quote

A customer texts a driveway photo. The estimator sizes it and reads condition; the learning engine prices it as a confidence band — not a wild guess — grounded in comparable jobs already in the books, benchmarked to the market, with the exact site factors that could move the number. A branded draft lands in the ledger with a measure-to-confirm clause.

× ft
autonomy with judgment

🔄 The autonomous lifecycle — with human money gates

One agent runs lead → quote → e-sign → weather-gated booking → FLHA → work → invoice → on-chain payment → review. It stops twice, both times about money. You are the human — clear the gates.

why on-chain, not a bolt-on

Programmable deposit → X Layer testnet

A blasting crew buys media and fuel before the first grain hits the driveway. The 25% deposit is enforced in code and encoded into an EIP-681 request — it clears in seconds (not a 3-day e-transfer), it's final (no chargeback after the abrasive is spent), and the agent verifies settlement itself by read-only RPC before it books the crew. Evolved never holds keys.

the asp earns too

🧾 x402 — pay-per-call, live

Watch the raw protocol: no payment → HTTP 402 with an accepts envelope; present proof → the service answers with a settlement receipt header, and the /stats scoreboard ticks.

6:30 am, rendered

🌅 The morning digest

Not JSON — the actual owner briefing, composed live from the books: the one thing not to drop, money pulse, today's jobs, blast-day weather.

the agentic cfo

📈 Should I buy a second truck?

A 12-month cash simulation from the live books: capex hole, winter trough, break-even month — drawn from the actual tool output.

pricing that learns

🧠 The rate engine — base market floor vs learned rate

Won jobs at healthy margins teach the engine — and it never stops. Every logged outcome lifts a confidence score, tightens the suggested quote range, and benchmarks the rate against the going market band. Watch what outcome history has already done to driveway pricing.

Adapt
Make it yours · one call

What business are you running?

Evolved is a toolkit, not a one-off. Pick a trade and preview exactly what franchise_spinup installs — the rate card into the quoting engine, the trade's hazards into every JHA, empty books, the full machine. Adding your own is one entry in src/trades.ts; brand_configure makes every rendered quote feel like your company. Read-only and safe to click here.

01
Pressure washing

I wash driveways & siding.

Rinse to strip-wash, priced per sqft. Wand-injection, slip, and ladder hazards in every JHA.

Preview the pack
02
Line painting

I stripe parking lots.

Re-stripe to full layout, per sqft. Live-traffic, fume, and heat-stress hazards baked in.

Preview the pack
03
Mobile detailing

I detail cars & fleets.

Express to full restoration, per unit. Chemical-exposure and customer-property hazards.

Preview the pack
04
Your trade

I run something else.

~30 lines in one file: your rate card + hazards. Then it is your OS — quoting, dispatch, payroll, on-chain invoicing.

See how
the toolkit story — try it

What franchise_spinup installs

Pick a trade above, or here, and preview exactly what re-seeds — rate card, depth labels, trade-specific hazards. The server also speaks the whole MCP spec: resources (rate table, hazard library, trade packs) and prompts, not just tools. Guide: docs/ADAPT.md · SECURITY.md

your books, your sheet

The workbook spine

The whole OS renders as a real operations workbook — every collection a tab. With a Google service account (EVOLVED_GOOGLE_SA) it creates, shares, and syncs an actual Google Sheets workbook; with zero credentials it exports the identical 20 tabs as CSV. Watch it write the spine live:

done work → next work

Scorecard & reputation

The Job P&L scorecard (quoted vs actual, win rate, overall margin) plus the reputation ledger — reviews earned, response rate, and the testimonial bank ready for a website.

power users

Point your own agent at it

Claude Desktop / Claude Code / any MCP client — the live Streamable HTTP endpoint:

{ "mcpServers": { "evolved": {
    "type": "http",
    "url": "https://www.evolvedmcp.cloud/mcp"
} } }
the full 83-tool surface

Or run it locally

Zero credentials; 41 tests including a live testnet probe:

git clone https://github.com/kr8tiv-ai/evolved.git
cd evolved && npm install && npm run build
npm test && npm run demo