Bike4Mind：开源核心的多模型 AI 知识平台

Your bicycle for the mind in the age of AI.

An open-core AI knowledge platform — a multi-model workspace where notebooks, autonomous agents, and a RAG knowledge engine work together to augment any cognitive task.

What is Bike4Mind?

Bike4Mind is an AI-powered knowledge platform that wires a superset of AI technologies together behind one workspace. Bring your own models, keep your data, and let agents do the heavy lifting.

Notebooks — the core workspace: chat, documents, and context in one place, across any model.
Multi-LLM by design — swap freely between Anthropic, OpenAI, Google Gemini, xAI, Ollama (local), and more, plus image models (FLUX, DALL·E 3, Stable Diffusion).
Quest Master — autonomous, multi-step task planning: text and image generation, vision review, web search, math, code, and human-in-the-loop steps, run in parallel.
AI Agents — autonomous ReAct-style agents that carry out complex tasks against your tools and knowledge.
Knowledge Engine — RAG over your documents: smart chunking, vector search, collections, and tagging so agents can reason about what you know.
Artifacts — reusable snippets, documents, and visualizations produced by agents, with a built-in publish-and-share layer.

Try the hosted service at app.bike4mind.com, or self-host the open core on your own infrastructure (below).

Open core

Bike4Mind is open core. The engine is public and self-hostable; the multi-tenant hosted service is our business.

Open (this repo, BUSL-1.1): the agent engine, LLM adapters, CLI, data models, and the self-host path — published as @bike4mind/* packages.
Commercial / closed: operating the multi-tenant hosted service — billing, entitlements, hosted infrastructure — and premium overlays such as Overwatch.

See CONTRIBUTING.md for the exact open/closed boundary.

Quickstart

Hosted (fastest): sign in at app.bike4mind.com — nothing to install.

Self-host: run the open core on your own hardware with Docker — no AWS account or cloud provider required. See the Self-Host Quickstart. In short:

# 1. Copy the env template, then generate JWT_SECRET / SESSION_SECRET /
#    SECRET_ENCRYPTION_KEY (see SELF_HOST.md) and add your model keys
cp .env.selfhost.example .env.selfhost

# 2. Bring up the stack (app + MongoDB + object storage + queues + mail catcher)
docker compose -f compose.selfhost.yaml --env-file .env.selfhost up -d

Bike4Mind then runs at http://localhost:3000; sign-in code emails land in the bundled Mailpit at http://localhost:8025. The self-host image is published to ghcr.io/bike4mind/bike4mind-selfhost.

Develop

Bike4Mind is a pnpm + Turborepo monorepo (Node 24).

pnpm i -r              # install
pnpm turbo:core:build  # build the @bike4mind/* core packages
pnpm turbo:typecheck   # type check
pnpm turbo:test        # run tests

Project layout: apps/client (Next.js SPA + pages API backend), packages/cli (interactive CLI + ReAct agent), b4m-core/* (the @bike4mind/* engine packages), packages/database (Mongoose models + migrations). Realtime WebSocket fanout ships as a separate @bike4mind/subscriber-fanout image. See CONTRIBUTING.md for the full guide.

Building an app that signs users in with Bike4Mind and bills AI usage to their accounts? See Building an app on Bike4Mind.

Contributing

Contributions are welcome — fork → topic branch → PR → squash-merge. All contributors sign a lightweight CLA on their first PR (you keep your copyright). Start with CONTRIBUTING.md and our Code of Conduct. Questions and self-hosting help go in Discussions.

Security

Please report vulnerabilities privately — see SECURITY.md. Do not open public issues for security problems.

License

Bike4Mind is licensed under the Business Source License 1.1 with a broad Additional Use Grant:

✅ You may read, modify, redistribute, and make production use of the code — including self-hosting it for your organization's own internal use, and building or commercializing your own products on top of it.
❌ You may not offer the software to third parties as a competing hosted/managed service (a "Bike4Mind Service", as defined in the LICENSE).
🕓 Each released version converts to Apache-2.0 two years after its public release. This license will never be tightened.

For alternative licensing, contact licensing@bike4mind.com.