§ GET STARTED · 3 MINUTES

Three minutes from sign-up
to your agent's first read.

Buy a book. Run one command. A fresh bundle of context — versioned, paid-for, hash-stored — unpacks straight into your agent's skills directory. Or skip the install entirely and connect your agent over MCP.

# Install gif-grep (free) — copy-paste, runs as-is on a fresh machine
$ mkdir -p ~/.claude/skills && curl -sL https://bkstr.tmrwgroup.ai/api/install/gif-grep | tar xz -C ~/.claude/skills/

# Or with the bkstr CLI (npm) — zero install via npx
$ npx -y @clawbot678/bkstr install gif-grep

Browse the catalog →Sign up free

VOL. 01EST. 2026STRIPE-BILLED · NO SUBSCRIPTION

§ WHAT IS BKSTR

A catalog of books your agent can read.

What bkstr is. A catalog of books — bundles of files your AI agent reads. Pay once per book. Fetch the files via API, install per your agent's docs, or connect over MCP.

What bkstr isn't. A prompt store. A no-code agent builder. A subscription. We don't run your agent for you; we don't see your conversations; we don't claim our books make any specific agent X% better. Quality varies by book and by use case.

Book

One purchase. Authored by a domain expert, versioned, hash-stored. Your access starts the moment Stripe webhook fires.

Files

The bundle — markdown chapters for books, executable files for skills. Delivered as a gzipped tarball that unpacks straight to disk.

Agent

Claude Code, Cursor, Cline, Aider — anything that reads files on disk, or anything that speaks MCP.

§ FAQ6 QUESTIONS

What's the difference between a book and a skill?

Both are bundles of files; the difference is how your agent uses them. Books are documents the agent reads as context — chapters of markdown, designed for grounded retrieval. Skills are files the agent installs and executes — a SKILL.md plus the scripts / configs / templates that make it run. The catalog treats them as one product type because the buy + fetch flow is identical; the install path differs by agent.

Can I use bkstr with [agent X]?

If the agent can read files from disk, yes. We document Claude Code, Cursor, Cline, and Aider explicitly under the Agent Developers tab above. For anything else, the pattern is the same: fetch the files JSON, write each entry to wherever the agent expects on-disk content (its rules folder, skills directory, project root, whatever the docs call it), and reference it from your agent's usual interface. Or — if the agent speaks MCP — skip the install path entirely and use the MCP tab.

How do book versions work?

Every book and skill has a version number — v1, v2, v3 — and the files endpoint returns the latest by default. Older versions are addressable via the ?version= query param. Authors commit to versioning discipline; breaking changes bump the major, copy fixes don't. Your purchase covers all future versions of the title you bought — re-fetch any time the publisher ships an update.

What if my API key leaks?

Revoke it from /dashboard/api-keys and issue a new one. Keys are hash-stored — bkstr never sees the plaintext after issuance, so a leak's blast radius is scoped to fetch access for keys that haven't been revoked. The revoke is immediate; any agent still holding the old key gets a 401 on next fetch.

How are books different from prompts?

A prompt is a single instruction you paste into a chat. A book is a curated bundle of files — versioned, paid-for, hash-stored — that your agent reads or installs as a durable artifact. Books carry their own chapter structure, sample inputs, and (for skills) executable scaffolding. The shorthand: a prompt is a sentence; a book is a chapter.

Is bkstr a subscription?

No. Each book or skill is a one-time purchase, billed through Stripe. Your access doesn't expire and re-fetches don't meter — pay once, fetch the latest version any time, forever. There's no plan tier, no seat math, and no auto-renewal.

§ READY?

Start with one book.