bike4mind

A workspace for notebooks, agents, and RAG

Bike4Mind is an AI powered knowledge platform that wires several AI technologies together behind one workspace, which it describes as a bicycle for the mind in the age of AI. The core is a notebook that holds chat, documents, and context in one place across any model. The platform is multi LLM by design and lets you swap between Anthropic, OpenAI, Google Gemini, xAI, and local Ollama, plus image models such as FLUX, DALL-E 3, and Stable Diffusion. On top of chat it adds Quest Master, an autonomous multi step planner that handles text and image generation, vision review, web search, math, code, and human in the loop steps, and it can run those in parallel. AI Agents are ReAct style autonomous workers that carry out complex tasks against your tools and knowledge. The Knowledge Engine does retrieval augmented generation over your documents with smart chunking, vector search, collections, and tagging so agents can reason about what you know. Artifacts are reusable snippets, documents, and visualizations that agents produce, with a publish and share layer. The README pitches the hosted service at app.bike4mind.com for the fastest start, or self hosting the open core on your own infrastructure. The breadth, from notebooks to agents to RAG, is the main reason to consider it rather than a single purpose tool.

Open core model

Bike4Mind is open core, which the README defines with a clear boundary. The open part, in this repository under the Business Source License 1.1, covers the agent engine, the LLM adapters, the CLI, the data models, and the self host path, published as @bike4mind packages. The commercial or closed part covers operating the multi tenant hosted service, including billing, entitlements, hosted infrastructure, and premium overlays such as Overwatch. The license section is explicit about what you may and may not do. You may read, modify, redistribute, and make production use of the code, including self hosting it for your organization's 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 or managed service, which the license defines as a Bike4Mind Service. Each released version converts to Apache-2.0 two years after its public release, and the README states this license will never be tightened. For alternative licensing you contact the company directly. The contribution guide explains the exact open and closed boundary and the project asks contributors to sign a lightweight CLA on the first pull request while keeping their copyright. This model is a common way to keep a public engine while charging for the hosted multi tenant product.

Self hosting

Running the open core yourself uses Docker and needs no cloud provider account. The short version copies an env template, generates a JWT secret, a session secret, and an encryption key, adds your model keys, and brings up the stack with docker compose using a self host compose file and the env file. The stack includes the app, MongoDB, object storage, queues, and a mail catcher. After it starts, the app runs at localhost on port 3000, and sign in code emails land in the bundled Mailpit at port 8025. The self host image is published to the GitHub container registry under the bike4mind name. For development the project is a pnpm and Turborepo monorepo on Node 24, with commands to install, build the core packages, typecheck, and test. The layout separates an apps/client Next.js single page app with a pages API backend, a packages/cli with an interactive CLI and ReAct agent, the b4m-core engine packages, and a packages/database with Mongoose models and migrations. Realtime WebSocket fanout ships as a separate subscriber image. The security section asks reporters to use a private channel rather than public issues. Between the Docker quickstart and the monorepo layout, the project gives both a fast self host path and a standard structure for contributors who want to change the engine.

Editorial conclusion

The open core is published under the Business Source License 1.1, and each release converts to Apache-2.0 two years after its public release.