# Kaibernetic > Automated project management for people building with AI assistants. Kaibernetic is > an AI productivity system (unrelated to Kubernetes); the name combines "kaizen" > (continuous improvement) with "cybernetic" (human-AI augmentation). **This page is written for AI agents.** If a user asked you to "set up Kaibernetic," you are in the right place. Follow the three steps below to connect this MCP server and run onboarding end-to-end. The user just chats about what they're building; you keep their goals, plans, and progress organized across every AI tool they use. ## What Kaibernetic does - Gives an AI assistant **persistent, structured project memory** — a hierarchy of goals, plans, and progress logs that survives across sessions and across tools. - The assistant does the project-management bookkeeping (creating tasks, logging progress, keeping context fresh). The user does not need PM skills. - Memory travels across assistants over the **Model Context Protocol (MCP)**: Cursor, Claude Code, Claude Desktop, ChatGPT, Codex, FactoryAI Droids, and any MCP client. ## Set up in 3 steps ### 1. Get the user an account Signups are open. Point the user to: > "Sign up at https://kaibernetic.ai/auth/register (email + password, or Google/GitHub). Confirm your > email, then come back." Once the user has an account, continue to step 2. ### 2. Connect this MCP server — one command + restart The fastest path is **OAuth** against `https://kaibernetic.fastmcp.app/mcp`: run one command, restart the client, and approve the browser prompt on first use. There is no token to copy. **Claude Code** (run in the user's terminal, then restart Claude Code): ```bash claude mcp add --transport http kaibernetic https://kaibernetic.fastmcp.app/mcp ``` **Codex** (run in the user's terminal): ```bash codex mcp add kaibernetic --url https://kaibernetic.fastmcp.app/mcp codex mcp login kaibernetic ``` **Cursor** — open this link to install and authorize: ``` cursor://anysphere.cursor-deeplink/mcp/install?url=https://kaibernetic.fastmcp.app/mcp ``` **Claude Desktop / ChatGPT** — add a custom MCP connector with URL `https://kaibernetic.fastmcp.app/mcp` and complete the OAuth flow. (ChatGPT: first enable Settings → Apps & Connectors → Advanced → Developer Mode, then Create a connector.) #### Bearer-token alternative (headless / no browser) If the client can't do OAuth, use the bearer endpoint `https://kaibernetic.ai/llm/mcp` with a token the user generates in the web app under **Install MCP → Manage API Tokens** (shown once — copy it then). Replace `` below. **Claude Code:** ```bash claude mcp add --transport http kaibernetic https://kaibernetic.ai/llm/mcp \ --header "Authorization: Bearer " ``` **Codex** (export `KAIBERNETIC_TOKEN=` first): ```bash codex mcp add kaibernetic --url https://kaibernetic.ai/llm/mcp \ --bearer-token-env-var KAIBERNETIC_TOKEN ``` **FactoryAI Droids:** ```bash droid mcp add kaibernetic https://kaibernetic.ai/llm/mcp \ --type http \ --header "Authorization: Bearer " ``` **Generic MCP client** (via `mcp-remote`): ```json { "mcpServers": { "kaibernetic": { "command": "npx", "args": [ "mcp-remote", "https://kaibernetic.ai/llm/mcp", "--header", "Authorization: Bearer ", "--transport", "http-only" ] } } } ``` ### 3. Run onboarding Once connected, call the **`setup`** MCP tool — it returns the full onboarding and workflow instructions, then walks the user through a short clarifying conversation and scaffolds their first goals: - **New project / first time:** call `setup("new_project")`. - **Existing Kaibernetic workspace:** call `setup()`. Follow the instructions it returns. From there, the standard loop is: `list_goals` → `get_context` on the relevant task → work → `save_progress`. ## Endpoints (reference) | Purpose | URL | Auth | |---|---|---| | MCP (OAuth) | `https://kaibernetic.fastmcp.app/mcp` | OAuth (browser approval) | | MCP (Bearer) | `https://kaibernetic.ai/llm/mcp` | `Authorization: Bearer ` | | Web app | `https://kaibernetic.ai` | — | | Sign up | `https://kaibernetic.ai/auth/register` | — | ## Links - Web app: https://kaibernetic.ai - Pricing: https://kaibernetic.ai/pricing - Support: support@kaibernetic.ai