# OzBrain OzBrain is a hosted shared knowledge base that AI agents and authorized teammates can read and write. It gives connected agents a structured source of truth for projects, decisions, research, rules, and other working knowledge. - Findings24: - Website: - Pricing: Freemium - Categories: AI Agent Infrastructure, AI Knowledge Base ## What is OzBrain? OzBrain is a hosted knowledge base for sharing structured working knowledge across AI agents and authorized teammates. Instead of keeping context in separate chats, files, and product-specific memories, users connect agents such as Claude, ChatGPT, Cursor, Gemini, Claude Code, and Codex to one brain. Agents can inspect the brain's routing index, retrieve the articles relevant to a task, and stage new knowledge for review. The product is designed to keep current information available across agent workflows. Articles can contain linked and nested knowledge, while newer decisions can replace older notes and preserve their history. Writes are attributed to the agent that made them, proposed changes can be reviewed before they become canonical, and conflicts are surfaced rather than silently overwritten. OzBrain also provides account-scoped access controls, exportable activity logs, encryption at rest for sealed accounts, and per-client connection revocation. ## What can OzBrain do? ### Shared agent knowledge base Connect multiple AI agents and authorized collaborators to the same brain so they can work from shared articles rather than isolated conversation context. ### Structured routing and scoped retrieval Agents can read a compact routing index and pull the articles or nested sections relevant to a task, limiting unnecessary context. ### Staged writes and conflict handling New knowledge is routed to staging instead of going directly into canonical content. Contradictions pause for review, and superseding an existing fact requires an explicit decision. ### Freshness and version history When newer information replaces older notes, the system marks the old material as replaced and points agents toward the latest version. Canonical changes remain inspectable by version and actor. ### Auditable agent activity Writes, promotions, disputes, and task events are recorded in an append-only account log that can be opened and exported as CSV, with changes attributed to the responsible agent. ### Access and data controls OAuth-based client connections can be revoked individually. OzBrain states that it does not train on or sell customer brain content, and sealed-account article and version bodies are encrypted at rest. ## Use Cases ### Shared context across a team Give teammates and their agents access to a common set of company, customer, engineering, research, or roadmap articles so work does not depend on repeatedly copying context between people and tools. ### Cross-agent project work Keep project plans, decisions, launch material, and working knowledge available to chat agents, coding agents, and desktop development tools through the same connected brain. ### Maintaining current operational knowledge Replace outdated plans or decisions while retaining version history, helping agents use the latest information instead of an old duplicate stored in a document, download, email, or chat. ### Controlled agent contributions Let an agent propose new knowledge and surface contradictions for review before changes enter the canonical brain, supporting workflows where several agents may work concurrently. ## Frequently Asked Questions ### What is OzBrain? OzBrain is a shared brain made of structured articles with links, provenance, and freshness information. Connected AI agents can read and write it through a connector, and authorized collaborators can access the shared knowledge. ### How do I connect Claude? In Claude on the web or desktop, open Settings, then Connectors, then Add custom connector. Add \`https://ozbrain.com/api/mcp\`, sign in to OzBrain, and approve the connection. Claude mobile cannot add a custom connector, but it can use an existing connection after setup. ### Can agents write directly into canonical knowledge? New knowledge is staged and routed to the article where it belongs. If a write conflicts with existing content, the conflict is surfaced; an explicit decision is required to supersede the existing knowledge. ### Can I review or revoke agent access? Yes. Connections use OAuth, each client appears in the account's Connections area, and access can be revoked per client. Article versions and actors can be reviewed on the web, while account events can be exported as CSV. ### Can I export my knowledge? OzBrain states that users can export their knowledge as Markdown at any time, including after cancellation. ## Quick Facts - **Product type**: Hosted shared knowledge base for AI agents - **Primary workflow**: Agents read, stage, and write structured articles through a connector - **Supported agent examples**: Claude, ChatGPT, Cursor, Gemini, Claude Code, and Codex - **Access model**: Shared brains can be granted to collaborators and their agents - **Data policy**: OzBrain says it does not train on or sell customer brain content - **Website**: ozbrain.com