Six beans · One library · No SaaS
A small-batch shop selling written instructions that turn AI coding agents into something resembling employees. Markdown brewed at 200°F, poured over your repo. Buy the bag, keep the bag. No "platform." No condescension.
Written and shipped by David Palmer in North Palm Beach, Florida. One operator, no support queue, and a reply address that reaches him. The measurement harness behind the memory work is public and MIT-licensed: the Memory Integrity Bench, if you would rather check the work than take the word.
"Money never sleeps. The library compounds. Neither does the espresso machine."
House motto · Unofficial · Unapologetic
One library, five levels of done. Read a skill free, buy a bean, take the whole board, have us cup your setup, or run the machines we run. Prices do not move while you scroll.
A working memory policy for your agent: what it stores, what it forgets, what it never mentions.
Your agent forgets your name between sessions and you call that "stateless architecture." This pack is the decision layer for what an agent remembers, forgets, and politely declines to mention. Episodic, semantic, procedural, and the policy discipline above all three. Stop letting your bot drown in its own state.
What's inside: 11 skills, 5 worked examples, and a runnable stack you can run with the supplied files. The recipes carry the write policy, ranking, per-user boundary, GDPR delete, and vector-index build order that silently returns nothing if you get it wrong. 27 files, 36,667 words. Every file listed in the pack's MANIFEST.
For: operators whose agent re-reads the repo every morning like it is new here. Bundle math: included in the Flight of Four ($599) and the Whole Board ($699), or $99 as the add-on for anyone who owns or is buying another pack.
The lane pattern for running several coding agents at once in separate git worktrees: how to split the work, gate the merges, and keep them off each other's files.
Make N AI agents do your job at the same time. They work in separate git worktrees so they don't punch each other, and they only reach main through a gate you define. This version of the pack is the method and the worked record of it: lane design, task specs, autonomy caps, the CI gate contract, and the merge discipline, with the real artifacts from a run. The automation scripts are included: 7 scripts ship in the archive; you wire the gate into your own CI and your own PR tooling, and you stay liable for the rebase. This is the actual pattern behind a multi-tenant SaaS built by one human operator who, statistically, was sleeping.
What's inside: 13 skills, 10 worked artifacts from real runs, and 7 scripts the pattern runs on. 50 files, 38,202 words, plus the test suite and install guide for Claude Code, Cursor, and Codex. Skill 1 is free to read right now, in full, before you spend anything.
For: a solo dev with a tested repo and a backlog wider than one keyboard. Bundle math: add ARL for $100 more as The Pair ($299), or take all four core beans for $599 in the Flight, $97 under à la carte.
Two agents tune your prompts, retrieval, or evals against a metric while you are elsewhere.
Two agents in a small room. One proposes. The other grades. They argue about prompts, RAG retrieval, evals, and copy until the numbers go up. You sleep. The bill goes up. The numbers go up faster. Convergence is the goal. Stopping conditions are included so you don't wake up to a $4,000 OpenAI receipt and a slightly better headline.
What's inside: 10 skills, 5 worked examples, and the loop itself: a runnable harness with SQLite persistence, resume-from-checkpoint, five stop conditions, and a 24-question retrieval benchmark you can run without an API key. 29 files, 20,062 words. Install guide for Claude Code, Cursor, and Codex.
For: anyone still tuning prompts by hand and calling it iteration. Bundle math: the other half of The Pair ($299 with PAL, $49 under separate bags). Want the grading rig prebuilt? The RAG Regression Harness waits in the Roastery.
Your product's API becomes an MCP server that agent clients can call in production.
Everyone is building a "GPT for X." You can do better. Turn your product's API into a production-grade MCP server: opportunity validation, tool schemas, auth, permission boundaries, prompt-injection threat model, Inspector testing, deploy paths, marketplace listing, monetization. Fourteen skills, end to end, with the schemas and threat model written out. No hand-waving.
What's inside: 14 skills across opportunity validation, tool schema design, auth and permission boundaries, a prompt-injection threat model, Inspector testing, deployment patterns and marketplace listing, plus 3 worked examples, 2 reference servers, and the container toolchain. 49 files, 28,301 words. Install guide for Claude Code, Cursor, and Codex.
For: product builders putting an API in front of Claude, Cursor, or any MCP client. Up the ladder: the Roastery sells this class of server already built ($1,499), and the Cupping Table will audit the boundaries you ship ($499).
One JSON manifest in, a consistent, metadata-clean WebP library out.
You need 80 on-brand images. Your design agency quoted $22,000. Midjourney quoted "good luck staying consistent across 80 prompts, buddy." This is a Python post-processing CLI plus the prompt discipline around it. You bring the rendered images and a JSON manifest; it gives back a metadata-clean, web-optimized WebP library, idempotently, on your own CPU, for nothing. Image generation is not in this version: the skills tell you how to prompt for consistency across 80 scenes, and the tool does everything after that.
What's inside: 8 skills, 3 example pipelines, 22 scene prompts, and the generation and post-processing layers. 37 files, 14,621 words. Run estimate before a prompt run to inspect backend cost.
For: anyone who needs 20 to 80 on-brand images without a design retainer. Bundle math: included in the Whole Board only ($699, $100 over the Flight); the $99 add-on below is for anyone who owns or is buying another pack, ordering à la carte.
Read more →The full Memory pack, $50 off, riding along with any other pack.
Own another pack, or buying one in this order? Add Memory at the regulars-only price. Fifty dollars off because you're not a tourist. (An agent without memory is a cause without consequences. We won't sell you that life.)
What's inside: identical to the standalone Memory pack. 11 skills, 5 worked examples, 27 files, 36,667 words.
For: à-la-carte buyers adding a second bag at checkout. Check your cart: Memory is already inside the Flight of Four and the Whole Board; if a flight is in your order, you do not need this.
All four core packs. Together. At a price your accountant will silently respect. Skip the bean-by-bean math: this is what the regulars buy on the first visit.
What's inside: 48 skills, 23 worked examples, 155 files, roughly 123,232 words across four zips.
Bean math: $599 across four beans is $149.75 a bag, and roughly 123,232 of the library's 137,853 words ride in this box. Only want two? The Pair is $299. Want the fifth bean too? The Whole Board adds the Image Pipeline for $100 more.
The Flight of Four is the house pour. If you want less, or you want the whole shelf, there is a bag for that too.
The shipping loop and the tuning loop, automated in one order.
Parallel Agent Lanes and the Auto-Research Loop, poured together. One runs your agents in parallel. The other makes two of them argue until the numbers go up. Buy the pair and the second cup costs less than the walk-in price.
What's inside: 23 skills, 15 worked examples, 79 files, 58,264 words.
For: operators automating both halves of the day: shipping and tuning. Upgrade math: adding MCP and Memory later comes to $647 by this route; the Flight of Four carries all four for $599.
Every bag on the shelf, one checkout, $139.80 a bean.
Everything on the menu that comes in a bag. All four core packs plus the Image Pipeline. The Flight of Four with the cold brew added. This is the shelf, not the sample.
What's inside: 56 skills, 23 worked examples and 3 example image pipelines, 192 files, 137,853 words, plus the Python CLI.
For: the operator who already knows the whole library is coming home. Bean math: the cheapest walk-in bag is $149; this prices all five at $139.80 each. Eyeing the Flight plus the Image Pipeline? Bought apart they ring up $748; this is $699.
Cupping is what a roaster does before they will vouch for a lot: sit down, grade it honestly, write up what is actually in the cup. Two audits. You buy, you tell us where to look, an automated pass runs against your target, and then an operator reads every finding before anything reaches you. The reviewed report lands within five business days of your intake. The beans above teach the patterns; the cupping table is us grading yours.
A map of where AI answer engines mention you, where they skip you, and who they name instead.
We ask the AI answer engines the questions your customers actually ask, in your service area, and record which of them cite you. An automated probe run, then an operator reads the findings before anything reaches you. Reviewed report within 5 business days of your intake.
For: brands whose buyers ask an AI before they ask a search engine. One engagement, one deliverable: you own the map when we hand it back. No retainer follows you out.
A written list of the ways into your agent or MCP server, and how to close each one.
An automated scan of your agent or MCP server: secrets, dependencies, static analysis, and your tool surface, followed by an operator review. You get every finding with the file and line it was found on. Reviewed report within 5 business days of your intake.
For: anyone shipping an agent or MCP server to people they have never met. Same boundaries as Bean IV: built your server with the MCP pack? This is the second pair of eyes on the same threat model: $199 to learn it, $499 to have it checked.
The packs teach your agents a pattern. These are the source: production-shaped templates and libraries you wire into your own stack and own outright. Heavier, built for the shop that runs its own line.
The class of server Bean IV teaches you to build, started for you in source.
An MCP server template with the auth already thought through. Cloudflare Workers, Hono, Drizzle on D1, and an OAuth 2.1 PKCE flow you wire to your own provider. You swap in your tools and deploy with Wrangler.
For: the shop that wants the head start, not the recipe. Down the ladder: the MCP pack ($199) teaches your agent to build one of these over your hours; this one starts from ours.
Governed decision memory, from local store to agent-facing MCP tools.
One 85-file private, single-user commercial source ZIP contains both TypeScript pieces: the @lonelybrewclub/memory-kit Control Plane and its @lonelybrewclub/memory-kit-mcp stdio server and tools. Use local SQLite or the included working Mem0 and Zep Cloud adapters, with search, delete, and capability reporting. Read governed current state and history; compile token-bounded Context Capsules with evidence, contradictions, and omissions; use the proposal-only legacy Markdown translator without changing canonical state.
For: Node.js 20+ agent projects that need inspectable current state, history, and context without flattening uncertainty. Install locally with pnpm. Down the ladder: Bean I ($149) teaches the pattern; this is the governed implementation in TypeScript source.
Metric drift fails the build before your users see it.
Stop shipping RAG changes on faith. A pytest CI gate for RAG repos: a schema-locked golden set format, a drift gate that fails the build when any metric drops more than 5 percent below your committed baseline, and a GitHub Actions workflow that posts a scorecard on every PR. Ships with deterministic offline scoring that you replace with your own evaluator.
For: teams whose RAG pipeline changes weekly and whose regressions currently ship. Down the ladder: Bean III ($149) runs tuning loops that need a finish line; this is the drift gate you wire into CI.
Everything above is for operators who want to run the machines themselves. This is for everyone else. You tell us what the business actually is, we design and ship the site, and then we keep it fed: hosting, certificates, backups, and something new on the page every month. One checkout covers both: $500 for the build plus $97 a month for the care that keeps it running. The full page on websites is here.
A complete site: copy, build, launch. Designed and shipped by hand, not assembled from a template you have already seen on four other businesses in the same zip code.
The site stays current without you thinking about it. Hosting, SSL, backups, an edit allowance, and one content post a month. The standing order that keeps the shop stocked after opening day.
One kinetic-typography ad, up to sixty seconds. Words timed hard to a beat, no stock footage, no voiceover you have heard before. Up to two per project; longer runtimes are quoted individually.
The website build is refundable in full until you approve the preview. The care plan cancels any month with no fee; the site and domain stay yours. Questions before you buy go to [email protected] and get a real answer, not a funnel.
Match the order to how much you want done for you. Never spent a dollar here? Read the free skill at /free. It is a complete file from the $199 Parallel Agent Lanes pack, and it will tell you whether the house style works on your stack before any money moves. One bean: pick the pain. Agents forgetting things is Bean I ($149). A backlog wider than one keyboard is Bean II ($199). Hand-tuning prompts is Bean III ($149). Shipping an MCP server is Bean IV ($199). Needing on-brand images in bulk is Bean V ($149). Exactly two beans: PAL plus ARL is The Pair at $299; any other two, buy the bags. Three or more: any three of the Flight's four beans lands $52 to $102 short of the Flight of Four at $599, so take the flight, and the Whole Board at $699 adds the Image Pipeline. Want it done for you instead? Audits at the Cupping Table run $499. Want our source to start from? The Roastery, $1,499 a machine. Run a local business and want the site built for you instead of the files? The website line at /sites is $500 to build and $97 a month to keep it running, and the site stays yours.
Mostly, and where it is not we say so. A skill is a plain text file your agent reads and executes, which means you can read the entire thing end to end before it ever touches your repo. Where a pattern needs code it ships as code you can read the same way: the Image Pipeline ships a Python post-processing program, and the Roastery tools are source from the first line. Every script takes --dry-run, every one reads a single config file you wrote, the pack's own test suite runs them against a throwaway repo with no network and no GitHub account, and nothing phones home. There is still nothing to audit that you cannot audit with cat. The value was never the file format. It is knowing which five prerequisites decide whether two agents can share a codebase, and what to do when the answer is no. We publish a complete skill from the Parallel Agent Lanes pack at /free, unedited and unlocked. Read it and decide for yourself.
Because a marketplace skill is one file and these are packs. Parallel Agent Lanes is 13 skills, 10 worked artifacts, and 50 files. The MCP pack is 14 skills across transports, OAuth, permission boundaries and a prompt-injection threat model, with 3 worked examples and 49 files. The comparison that matters is not one file against another file. It is the pack against one hour of a senior developer's time, which is what you will spend rediscovering the third prerequisite the hard way. Buy once, keep the files, v1.x updates are free for life.
More than most shops will let you. Every product lists its full manifest right on this page: exact skill count, worked-example count, file count, word count. One complete skill is published free at /free, the same file buyers get, nothing held back. Every pack ships a MANIFEST describing each file it contains. Everything is plain text. And if the zip still is not what you expected, reply to the delivery email with the word "refund" inside seven days and we send your money back without an interview.
A folder of markdown files, worked examples, and where the pattern needs it, scripts. You drop them into your agent's context: Claude Code skills, Cursor rules, Codex AGENTS.md, any agent that loads a .md file. The pack teaches the pattern. The agent applies it. You own the files forever. Packs and Roastery tools are bought once and owned. The only recurring line in the shop is the website Care Plan at $97 a month, and it cancels any month with no fee.
For the packs and the Roastery tools: because we already wrote the thing. You're going to pay us. The transaction is done. No monthly tax, no "your free tier is over" emails at 11pm, no surprise pricing changes when our investors demand growth. Buy once. Own forever. Sync to whatever machine you want. The website Care Plan is the one thing we bill monthly, because keeping a site running is monthly work; it cancels any month and the site and domain stay yours.
Seven-day refund, no interview, no exit survey. Reply to the delivery email with the word "refund" and we send your money back. After day seven we still listen, but the receipt is closed. The website build is refundable in full until you approve the preview. The care plan cancels any month with no fee; the site and domain stay yours.
v1.x updates are free for life. New skill variant, bug fix, better example, fresh stack pattern: existing buyers get it. We'd only charge for v2.0 (significant restructure), and v1.x owners get a discount. We're not running a Subscription as a Service.
For the packs, yes: build with them, ship what you build, charge what it is worth. The line is redistributing the files or wrapping them into a competing product or course. Roastery tools are single-user source: deploy what you build for a client, but the source and the tool stay with you, and a second seat is a second license.
There are exactly seven Kybalion hermetic principles. We name the roasts after them, and we are still a bean short, which we have decided to call a roadmap. Hermes also happens to be the classical patron of merchants, messengers, and language: the literal job description of "a store that sells written instructions." We're not running a cult. We're running a coffee shop that read too much philosophy.
Because solo operators with AI agents are the new espresso machine, the new home roastery, the new "I do this myself, slowly, and it tastes better than the chain." That's the whole product. The metaphor and the offering rhyme.
You buy the audit like anything else on the menu. Right after checkout you land on a short intake form where you tell us where to look: your brand and competitors for the GEO Audit, or your repo or endpoint and scope for the Security Audit. We run a single pass, review the findings ourselves, and send the report to your purchase email within five business days. One engagement, one report, no retainer.
A pack is a set of markdown instructions that teaches your agent a pattern; you keep the files and the agent does the work. A Roastery tool is the source itself: code you wire into your own stack, like the MCP server template, the Memory Kit library, or the RAG harness. Same shop, heavier gear. You own the source outright, single-user commercial license, same 7-day refund.
Because they sit at opposite ends of the counter. A pack is instructions: your agent builds the thing across your hours and your API bill, and you own what it builds. A Roastery tool is the thing itself, in source: code you clone, wire to your own accounts, and own outright, single-user commercial license, same 7-day refund. You are paying for the distance between a recipe and a crated machine on your bench. If that distance is not obviously worth it for your shop, buy the bean and build it yourself. That is a fine outcome, and we sell both paths on purpose.
Optional · the free skill stays free either way
We will send the full contents list of the Parallel Agent Lanes pack, plus a note when a pack updates. Nothing else, and one click unsubscribes. Confirming is a second click, on purpose.