How much does Bike4Mind cost?

Individual is $15/month and includes $10 of usage as a transparent Balance (real dollars, not credits). Team is $20 per user/month with a shared Balance pool. Enterprise is a custom flat license deployed in your own AWS, with bring-your-own model keys at 0% markup. There is no free tier.

Does Bike4Mind use credits or dollars?

Dollars. Bike4Mind retired credits in favor of a transparent Balance shown in real dollars — every model and API call shows its actual cost, you set a hard spend ceiling, and the markup falls as you scale (20% to 10% to 5%), or 0% if you bring your own keys.

Is there a free tier or free trial?

No free tier — by design. Bike4Mind focuses entirely on paying customers so resources aren't spent chasing freeloaders. Agents and bots are welcome to try (see bike4mind.com/agents); production use just needs a key. Individual access starts at $15/month, and you can run real code with no account on the in-browser Python demo at bike4mind.com/features.

Can I self-host Bike4Mind / does my data leave?

Yes. Bike4Mind deploys in your own AWS account with full source. Your data never leaves your cloud, you hold the encryption keys, and inference can run on AWS Bedrock or self-hosted open-weight models so no third-party model vendor is ever in the data path.

Is Bike4Mind open source?

Bike4Mind is open core: source-available today under FSL 1.1, converting to Apache-2.0 in two years, locked by a published covenant so it cannot be quietly relicensed. You can read it, run it, and self-host the whole thing.

Can I bring my own API key (BYOK)?

Yes — bring your own model provider keys and pay 0% markup. Use your keys or ours; the more you use, the lower the rate.

What AI models does Bike4Mind support?

60+ models across OpenAI, Anthropic, Google, xAI, and Meta, plus any open-weight model you self-host via Ollama. Switch between them mid-conversation through one interface.

Is Bike4Mind HIPAA, SOC 2, and GDPR compliant?

Bike4Mind is architected for regulated workloads: deployed in your own AWS (or GovCloud), PHI never leaves your cloud, your KMS keys, full audit logging, and no prompt or response retention. Compliance is operated as a continuous program; request the security package for BAA terms and current SOC 2 status.

Does Bike4Mind have an API and a CLI?

Yes. There is a REST API, a published command-line tool (npm i -g @bike4mind/cli), and MCP support as both host and server. See bike4mind.com/for-developers and /agents.

Can I import my ChatGPT and Claude history?

Yes — one-click import of your ChatGPT and Claude conversation history.

Open by contract

Your AI keeps running when theirs doesn't.

The full AI workbench — chat, agents, voice, knowledge, images — that runs any model, from OpenAI and Anthropic to open-weight on your own hardware. Start hosted in one click, or run the whole thing inside your own AWS. When a vendor hikes prices, revokes your key, or kills a model, you don't even flinch.

✓
Swap models without re-architecting — no vendor in the middle
✓
When a model gets cut, deprecated, or repriced, you keep serving
✓
Your data never leaves your AWS

Get started See how it works

Open core

Run the whole thing yourself. Source-available today under FSL 1.1, real Apache-2.0 automatically in two years — the no-rug-pull clause is written into the license itself.

View the source

Enterprise by subtraction

Tell us what you don't need — that's how we build your platform. Chat, agents, voice, CLI, RAG, collaboration, images — all included. You decide what ships.

The neutral runtime

A lab can't be model-neutral.
A cloud can't be cloud-neutral.

We can be both — that's the whole point. Bike4Mind is the orchestration layer that treats every model as a commodity: frontier APIs when they're ahead, open-weight models on your own hardware when it matters that nobody can take them away.

One workbench, one API call. The frontier labs are interchangeable peers — the self-hosted lane is the one thing no lab or cloud can ever take away.

Every frontier model, one workbench

60+ models across OpenAI, Anthropic, Google, and AWS Bedrock behind a single interface. No lab will ever route you to its rival's better answer. We will.

Open-weight on your own iron

Qwen, Llama, DeepSeek — self-hosted via vLLM, on hardware you control. The one lane no lab can sell you and no directive can switch off.

Swap mid-conversation

Your sessions, prompts, artifacts, and data lakes don't care whose model answered. When a provider gates its best tier or reprices overnight, you change one dropdown — not your architecture.

One Pane of Glass

Notebooks, agents, artifacts, images, data lakes — the whole workbench, one login, any model. The pieces you'd otherwise cobble together, already speaking to each other.

Artifacts — 13 Types

Live React components, in-browser Python, Mermaid, charts, even chess boards — versioned like code, Claude-Artifacts-compatible, and rendered in a locked-down sandbox on your machine.

Agents & Quests

Long-running agentic jobs with real tools — code REPL, search, MCP — under hard budget guards. Fire a quest, come back to finished work.

Any Model, One Key

60+ models across OpenAI, Anthropic, Google, and Bedrock behind one workbench. Switch instantly, mid-conversation. Your balance rolls over for 3 months.

Open-Weight on Your Iron

Self-host Qwen, Llama, or DeepSeek via vLLM for complete privacy and control — the lane no frontier lab will ever sell you.

Data Lakes & Knowledge

DOCX, PDF, images, code files, and more — vectorized, searchable, and retrieved into any session with advanced RAG.

Multiplayer Notebooks

Shared sessions and team workspaces — collaborate inside the same conversation, not around it. Enterprise SSO and role-based permissions included.

No login · no server · runs on your machine

Don't take our word for it. Run it.

Two live artifacts, right here on this page: real Python executing via WebAssembly, and a real React component transpiled and rendered in a locked-down sandbox. In the product these are two of thirteen artifact types — Claude-Artifacts-compatible, versioned like code, and executed on your machine, never on our servers. Edit either one and run it again. This page never phones home.

artifact type: python — a tiny language model, trained live

shakespeare.py

# Train a tiny language model on Shakespeare — live, in YOUR browser.
# This is real Python, running on your machine via WebAssembly.
# No server. No login. No API bill. Edit me and run again.
import random
random.seed(7)

text = """To be, or not to be, that is the question:
Whether 'tis nobler in the mind to suffer
The slings and arrows of outrageous fortune,
Or to take arms against a sea of troubles
And by opposing end them. To die, to sleep,
No more; and by a sleep to say we end
The heart-ache and the thousand natural shocks
That flesh is heir to."""

ORDER = 4  # try 2 (chaotic) or 6 (smarter) and run again

# "Training": learn which characters tend to follow each context.
model = {}
for i in range(len(text) - ORDER):
    ctx, nxt = text[i:i + ORDER], text[i + ORDER]
    model.setdefault(ctx, []).append(nxt)

# "Generation": sample new Shakespeare-ish text, one character at a time.
ctx = text[:ORDER]
out = [ctx]
for _ in range(320):
    nexts = model.get(ctx)
    if not nexts:
        break
    c = random.choice(nexts)
    out.append(c)
    ctx = (ctx + c)[-ORDER:]

print("".join(out))

output

Press Run. The first run downloads the Python runtime once (~6 MB from a public CDN), then it's instant.
Your code and output stay on this page — they are never sent to a server.

Powered by Pyodide (CPython compiled to WebAssembly). The same model-agnostic, runs-where-your-data-is philosophy underpins the whole platform.

artifact type: react — a live component, transpiled in your browser

No login · no server · runs on your machine

Now do the same thing with React.

This is a real React component, transpiled by Babel and rendered live in your browser — inside a sandboxed, opaque-origin iframe that cannot phone home. It's the same artifact sandbox that ships inside Bike4Mind. Edit the loot table, press Run, and roll.

loot-drop.jsx

// Live React — transpiled and rendered in YOUR browser.
// No build step. No server. Edit me and press Run.
const { useState, useRef } = React;

// --- Knobs: tweak these and hit Run ---------------------------------
const RARITY_COLORS = {
  common:    "#9aa4b2",
  rare:      "#3b82f6",
  epic:      "#a855f7",
  legendary: "#f59e0b",
};

// A d20 roll (1-20) decides what drops. Higher rolls = shinier loot.
const LOOT_TABLE = [
  { min: 1,  rarity: "common",    items: ["Rusty Dagger", "Cracked Shield", "Stale Bread"] },
  { min: 11, rarity: "rare",      items: ["Silver Ring", "Oaken Longbow", "Healing Draught"] },
  { min: 17, rarity: "epic",      items: ["Stormcaller Axe", "Cloak of Whispers"] },
  { min: 20, rarity: "legendary", items: ["Dragonheart Blade", "Crown of the Sky"] },
];
// --------------------------------------------------------------------

const pick = (a) => a[Math.floor(Math.random() * a.length)];
const d20 = () => 1 + Math.floor(Math.random() * 20);

function lootFor(roll) {
  const tier = [...LOOT_TABLE].reverse().find((t) => roll >= t.min);
  return { rarity: tier.rarity, item: pick(tier.items) };
}

function App() {
  const [roll, setRoll] = useState(null);
  const [drop, setDrop] = useState(null);
  const [rolling, setRolling] = useState(false);
  const [bag, setBag] = useState([]);
  const [burst, setBurst] = useState(0); // bump to fire confetti
  const timer = useRef(null);

return (
    <div className="stage">
      <div className="die-wrap">
        <div className={"die" + (rolling ? " rolling" : "")} style={{ borderColor: color, color }}>
          {roll ?? 20}
        </div>
        {burst > 0 && (
          <div className="burst" key={burst}>
            {sparks.map((s, i) => (
              <span className="spark" key={i} style={{ "--tx": s.tx, "--ty": s.ty, background: color }} />
            ))}
          </div>
        )}
      </div>

<button className="roll-btn" onClick={rollLoot} disabled={rolling}>
        {rolling ? "Rolling…" : "Open Loot Drop"}
      </button>

<div className="drop" style={{ color }}>
        {drop ? drop.rarity.toUpperCase() + " — " + drop.item : "Roll the d20 for loot!"}
      </div>

<ul className="bag">
        {bag.map((it) => (
          <li key={it.id} style={{ color: RARITY_COLORS[it.rarity] }}>
            <b>{it.roll}</b> {it.item}
          </li>
        ))}
      </ul>
    </div>
  );
}

previewruns on your machine · no server

How the sandbox works: artifacts render in an opaque-origin iframe (sandbox="allow-scripts", never allow-same-origin) with connect-src 'none'— artifact code can render anything and exfiltrate nothing. When the repository opens, read the sandbox source yourself.

A Guild of Raiders, Not a Lobby of AFK Players

We spent three years invite-only, building for paying customers with zero ad spend — bootstrapped, profitable, shipping weekly. Now the doors are open. Same guild, same standards, room for you.

Doors Open

Three years invite-only, built for paying customers. Now the invite wall is down and the source is opening. Walk in, read the code, fork it if we ever disappoint you.

Bootstrapped & Profitable

No investor pressure to grow vanity metrics. Every dollar goes into features our customers actually need and use. Three years in production.

Builders, Human & Agent

Serious builders — people and agents — and the entrepreneurs and enterprises who value excellence. Agents are first-class citizens here.

Honest Credits

Pricing that maps to real compute. Your balance rolls over for 3 months. No bait-and-switch. No feature gates.

A note from the founder

"I've been shipping software for thirty years — from Starfleet Command to FarmVille, MMOs to virtual worlds. I'm not a frontier-lab insider. I'm a builder who watched the leverage of AI concentrate into a handful of companies and decided a counterweight had to exist. Bike4Mind is that bet: strong AI and agentic systems, open by contract, for everyone. Fork it. Build on it. Prove us honest."

— Erik Bethke, founder

Four Products. One Team. Zero Excuses.

The same engine that powers Bike4Mind is already running BedrockNews, StocksandVibes, and K2Kanji in production. This is not a demo. This is not a roadmap. These are live URLs you can click right now.

bedrocknews.com

BedrockNews

News Without the Negativity

Journalism that proves you can be truthful AND uplifting. Real reporting, real sources, zero rage-bait.

AI-Curated

Real-Time

Early Access

Visit BedrockNews

app.bike4mind.com

Bike4Mind

The Cognitive Workbench

60+ AI models, RAG pipelines, multi-agent collaboration, voice, artifacts, and Quest Master — deployed in YOUR AWS.

60+ Models

Enterprise Ready

Live Now

Open the Workbench

stocksandvibes.com

StocksandVibes

Trading Education in Plain English

Describe a trade in plain English. Get four strategies back. Paper trade with $100K virtual funds. Learn without the risk.

$100K Paper Trading

4 AI Strategies

Live Greeks

Visit StocksandVibes

k2kanji.com

K2Kanji

Thermal Learning for Kanji

Master Japanese kanji — meaning, reading, writing, and listening — with an adaptive “thermal learning” engine that keeps you in flow. The same method extends to Chinese, Korean, and beyond.

Thermal Learning

Adaptive Levels

Live Now

Visit K2Kanji

B4M API

When one API can power AI journalism, trading intelligence, language learning, and a cognitive workbench, you know the infrastructure is real. Your deployment won't be the first thing built on this engine. It'll be the fifth.

Ship Log

672 prod deploys in 925 days — roughly 5 per week.

Jun 29

Data lake management, smarter knowledge base surfacing, and more reliable artifacts

Data Lake Management Panel

Drag-and-Drop Ingest from the Explorer

Jun 28

Smarter optimization controls, artifact improvements, and more reliable solver feedback

Optimization Map Controls & Active Brief

Popped-Out Lens Fills Modal

Jun 27

Smarter agent mentions, faster cohort analytics, and a more reliable login experience

Tavern Agent Workstations

Reliable Agent @Mentions

Jun 26

Collaborative annotations, AI-powered revision, instant takedowns, and a refreshed profile experience

Collaborative Annotation Layer

Instant Content Takedowns

See the full changelog

Professional AI Workspace

Projects, RAG, and multiplayer sessions in one workspace — built for developers, research teams, and enterprises.