roastify-mcp
MCP server for the Roastify Developer API, letting AI agents browse the coffee catalog and generate packaging artwork, with paid tools settled in Bitcoin Lightning.
README
roastify-mcp
A Tollbooth DPYC operator for the Roastify Developer API. Roastify is a white-label coffee dropshipper: it roasts, prints branded packaging, and ships direct to customers.
This operator lets an AI agent read the Roastify catalog and generate packaging artwork, paid for in Bitcoin Lightning satoshis with no account, no KYC, and no subscription.
What you're building
YOU (operator, human in the loop)
│ │
│ set & tune prices │ drive credential intake
▼ ▼
┌───────────────┐ ┌──────────────────────────────────────────────┐
│ Pricing Studio│ prices │ Roastify — YOUR OPERATOR MCP │
│ (iOS) ├───────▶│ FastMCP · deployed on Horizon │
└───────────────┘ Neon │ ┌────────────────────────────────────────┐ │
│ │ roastify domain module ◀─ Roastify API │ │
Patron (Citizen) │ │ @runtime.paid_tool(FROZEN_UUID) tools │ │
+ MCP client ─────────▶│ ├────────────────────────────────────────┤ │
(Claude, etc.) npub │ │ tollbooth-dpyc SDK (the wheel) │ │
+sats │ │ ledger · vault (AES-256-GCM) · pricing │ │
│ │ /ConstraintGate · Secure Courier·audit │ │
│ └────────────────────────────────────────┘ │
└───┬─────────┬──────────┬───────────┬─────────┘
▼ ▼ ▼ ▼
Neon Postgres BTCPay▶ Sponsor Nostr relays
(your schema) Lightning Authority proofs·courier
ledger+pricing invoices certify + DMs·audit
provision │
api.roastify.app ◀── domain │ ▼
(each patron's calls └──▶ DPYC Oracle +
own API key) dpyc-community
Onboarding roadmap
- Nostr keypair — generate one (
nak key generate). The nsec is the only env var:TOLLBOOTH_NOSTR_OPERATOR_NSEC. - Sponsor Authority — register; it provisions your Neon schema automatically.
- Secure Courier — deliver your BTCPay secrets (
btcpay_host,btcpay_api_key,btcpay_store_id) by Nostr DM. They are vaulted, never put in the environment. - Set prices in Pricing Studio — see below. Tools start unpriced and uncallable.
- Deploy on Prefect Horizon —
.fastmcp.yamlis ready.
Get Pricing Studio
Your tools ship unpriced, and nobody can call a paid tool that has no price. Pricing Studio (iOS) is the operator console: it reads and writes the pricing model live in Neon, so prices never live in code — see tollbooth-pricing-studio.
That indirection is the point. Dynamic per-tool pricing — surge, happy-hour, loyalty discounts, free trials for a first-time patron — is what DPYC gives you that a flat paywall never could.
Bring your own Roastify key
Every patron delivers their own Roastify API key through the Secure Courier, vaulted per npub. There is no operator-held key and no fallback path.
This follows from the domain rather than being a policy choice. Roastify scopes the catalog, saved designs, and plan tier to the merchant account behind the key — a patron on Base sees a different catalog than one on Pro. A single operator-held key would hand one merchant's world to every caller, which is a wrong answer rather than a limitation to work around. It also means this operator never holds a credential it could misuse, and revoking your key in the Merchant App severs access without involving the operator.
roastify_get_patron_onboarding_status # what's missing
roastify_request_patron_credentials # → you get a Nostr DM
(reply with your Roastify API key)
roastify_receive_patron_credentials # vaulted, encrypted at rest
Live keys start rty_live_. Sandbox keys start rty_test_ and do not fulfill — useful for
exercising the read tools without touching production.
Tools
| Tool | Tier | What it does |
|---|---|---|
roastify_browse_catalog |
read | Catalog products and coffee blends together, plan tier marked |
roastify_get_catalog_product |
read | One product with variants, prices, stock, dieline template |
roastify_get_blend |
read | One blend: roast level, decaf status, variants |
roastify_list_my_products |
read | Your saved designs, cursor-paginated |
roastify_get_my_product |
read | One saved design in full |
roastify_check_stock |
read | Stock for one SKU or the whole list |
roastify_generate_artwork |
write | Generate artwork from a saved design |
roastify_artwork_status |
free | Check that job |
Plus the standard DPYC catalog (balance, purchase, courier, pricing, Oracle, status).
What this operator does not do, and why
It cannot create, update, delete, or sync products. Not a missing feature — the Roastify
API has no such endpoints. /v1/products is read-only, and the words integration, sync,
shopify, and upload do not appear anywhere in the
OpenAPI spec. Saved products and storefront sync are
Merchant App capabilities by design.
It does not place orders. Roastify already routes orders to its own operations and reports tracking back to your storefront. An MCP layer between those adds a hop without adding judgment, so order automation is deliberately left to Shopify and Roastify.
Artwork generation is template personalization, not design. roastify_generate_artwork
rewrites the named text and image placeholders of a design you already authored in the
Roastify Design Studio. It cannot author a design from scratch.
A generated artwork URL is not attached to anything. In Roastify's API the only consumer of an artwork URL is an order line — and orders are out of scope here. You receive a URL and carry it onward yourself. That seam is accepted deliberately; it is not an oversight, and this operator will not simulate a design flow the API cannot complete.
The catalog has no palate data. Origin, region, altitude, processing method, and varietal
exist only in the Merchant App UI. The machine-readable signals are roastLevel, isDecaf,
size, and price; everything else about a coffee's character is prose in its description. Bean
selection is correspondingly weaker than it looks, and this operator does not dress up an
LLM's reading of a description as structured provenance data.
Artwork is asynchronous
Roastify renders artwork on its own servers, so roastify_generate_artwork hands back a job
id straight away and roastify_artwork_status checks it. Status checks are free — you
learn the work finished by looking, and charging for each look would be charging for waiting.
job = roastify_generate_artwork(product_id=..., fields=[...], client_req_id="my-req-1")
roastify_artwork_status(job_id=job["job_id"]) # free, poll until artwork_url appears
Pass client_req_id and a repeated request is safe: it becomes Roastify's
Idempotency-Key, so the artwork is not generated twice.
The operator does no waiting of its own — both calls return immediately — so there is no detached executor here and nothing to recycle. That machinery would earn its place only if this operator ever rendered artwork itself.
Development
pip install -e ".[dev]"
ruff check .
pytest -v
python -m roastify_mcp.server # needs TOLLBOOTH_NOSTR_OPERATOR_NSEC to fully run
Upstream status
Roastify's API is version 0.3.1 and its own spec marks it beta, "subject to change." Expect to follow breaking upstream edits, and weigh that before promising this operator's catalog to third-party merchants.
License
Apache 2.0 — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.