# Remnus > Remnus is the Human-Agent Collaborative Workspace — pages and databases that Claude, Cursor, and any MCP-compatible AI agent can read and write over a standard HTTP API, under the same audit trail your team sees. Remnus pairs a page-and-database editor (pages, databases, kanban/table/calendar views) with a built-in Model Context Protocol (MCP) server, so an AI agent can search, read, and write workspace content directly instead of copy-pasting through a chat window. ## Wiki MCP reference documentation — endpoint, authentication, tools, resources, and prompts. - [MCP Documentation](https://remnus.com/wiki): Remnus MCP is a Model Context Protocol server built into every Remnus workspace. It lets AI agents — Claude, Cursor, Windsurf, and any MCP-compatible client —… - [Getting Started](https://remnus.com/wiki/getting-started): This guide walks you through connecting your first AI client to Remnus. The recommended path is OAuth — no token to generate or paste. A personal access token… - [Connect Your Editor](https://remnus.com/wiki/connect-editors): Remnus exposes a single remote MCP endpoint that any MCP-compatible editor can connect to: - [Authentication](https://remnus.com/wiki/authentication): Remnus supports two ways for an agent to authenticate to the MCP endpoint: OAuth 2.1 and personal access tokens . Both are workspace-scoped — a single token… - [Read Tools](https://remnus.com/wiki/read-tools): All 11 read tools are available to every token regardless of scope. - [Write Tools](https://remnus.com/wiki/write-tools): Write tools require a write-scoped token. Calling these with a read-scoped token returns an error and makes no changes. - [Resources](https://remnus.com/wiki/resources): MCP resources provide structured, addressable data that clients can subscribe to or read on demand. Remnus exposes six resource templates. - [Prompts](https://remnus.com/wiki/prompts): MCP prompts are reusable templates that fetch workspace data and return a filled message ready for LLM completion. The LLM call itself is performed by the… - [Agent Memory](https://remnus.com/wiki/agent-memory): Agent Memory turns a Remnus workspace into a durable, human-readable memory your AI agent can save to and recall from over MCP — decisions, preferences,… - [Context-First MCP](https://remnus.com/wiki/context-first): Remnus uses Open Knowledge Format ideas to make workspace knowledge portable, attributable, reviewable, and easier for AI agents to retrieve. OKF is the… - [Token-Efficient Usage](https://remnus.com/wiki/token-efficient-usage): Every MCP read costs an agent tokens, latency, and context-window space. Remnus exposes the workspace's existing structure — the tree, database schemas, the… ## Docs Articles on the project's architecture, licensing, and build decisions. - [MCP Security Guide: OAuth, Tokens, Scopes, and Audit Logs](https://remnus.com/docs/mcp-security-guide): A verified guide to MCP security: the threat model (malicious servers, prompt injection, excessive permissions, credential leakage), OAuth and PATs, least privilege, destructive-action confirmation, audit logs, and how Remnus applies each control. - [How to Run Headless AI Agents Without Login Prompts](https://remnus.com/docs/run-headless-ai-agents-with-mcp): A practical, verified guide to authenticating headless AI agents against a Remnus workspace: OAuth vs personal access tokens, secure token storage, minimum-permission scopes, and a real headless Claude Code example. - [AI Agent Memory vs RAG vs Context Windows](https://remnus.com/docs/ai-agent-memory-vs-rag-vs-context-window): A technical comparison of context windows, retrieval-augmented generation, and persistent agent memory: primary-source definitions, a direct comparison table, and where to store what in a real project. - [Why Coding Agents Forget Your Project Between Sessions](https://remnus.com/docs/why-coding-agents-forget-your-project): Why Claude Code, Cursor, and Codex lose project context between sessions, and how instruction files, documentation, task tracking, and memory combine into a context system that actually persists. - [How Remnus Uses OKF to Give AI Agents Better Context](https://remnus.com/docs/okf-context-engine-for-ai-agents): How Remnus combines portable OKF knowledge, exact-revision human review, Context Pack v2, and Smart or Strict MCP policies for AI agents. - [How to Let an AI Agent Manage Your Kanban Board](https://remnus.com/docs/ai-agent-manage-kanban-board): Learn how to use an AI kanban board safely with agent-assisted triage, task updates, human approvals, MCP workflows, prompts, and failure controls. - [AI Agent Project Management: The Complete Guide](https://remnus.com/docs/ai-agent-project-management-guide): A practical guide to AI agent project management: realistic responsibilities, human oversight, project lifecycles, safeguards, and MCP-enabled workflows. - [How to Connect OpenAI Codex to Remnus with MCP](https://remnus.com/docs/connect-openai-codex-to-remnus-mcp): Connect OpenAI Codex to a Remnus MCP workspace for requirements, tasks, decisions, documentation, status reports, and durable agent memory. - [How to Connect Cursor to Remnus with MCP](https://remnus.com/docs/connect-cursor-to-remnus-mcp): Connect Cursor to a Remnus workspace over MCP: verified mcp.json setup, OAuth vs personal access tokens, global vs project config, five workflows, and troubleshooting. - [What Is an MCP-Native Workspace? A Complete Guide](https://remnus.com/docs/what-is-an-mcp-native-workspace): An MCP-native workspace is built for AI agents from the start: readable, writable, scoped, and audited over the Model Context Protocol. Here is how the model works and where it fits. - [How to Connect Claude Code to Remnus with MCP](https://remnus.com/docs/connect-claude-code-to-remnus-mcp): A verified, step-by-step guide to connecting Claude Code to a Remnus workspace over MCP — OAuth vs personal access tokens, scopes, first workflows, and troubleshooting. - [How to Give Claude Code Persistent Memory and a Shared Workspace](https://remnus.com/docs/claude-code-persistent-memory-workspace): Learn how to combine Claude Code memory, CLAUDE.md, project tasks, and an MCP workspace for reliable long-running software work. - [How Many Tokens Does Your Agent Burn Reading Your Notes?](https://remnus.com/docs/agent-token-efficiency): A measured look at agent workspace reads, plus how projection, outlines, digest, delta sync, and OKF-aware Context Pack v2 control context cost. - [Remnus vs Notion: Full Comparison](https://remnus.com/docs/remnus-vs-notion-mcp): An honest, fact-checked look at Remnus vs Notion: workspace features, MCP authentication, agent tooling, pricing tiers, and audit trails. - [Remnus vs AppFlowy: Full Comparison](https://remnus.com/docs/remnus-vs-appflowy): A fact-checked comparison of two AGPL-3.0 Notion alternatives: AppFlowy's mature editor versus Remnus's first-party MCP server and agent tooling. - [Remnus vs AFFiNE: Full Comparison](https://remnus.com/docs/remnus-vs-affine): A fact-checked comparison of AFFiNE's mature editor and offline-first design against Remnus's first-party MCP server and open self-hosting. - [Remnus vs Obsidian: Full Comparison](https://remnus.com/docs/remnus-vs-obsidian): A fact-checked comparison of Obsidian's local-first notes and plugin ecosystem against Remnus's first-party MCP server for team workspaces. - [How I Built Remnus, an MCP-Native Open-Source Workspace](https://remnus.com/docs/how-i-built-mcp-native): The build story, from the headless agent that Notion's MCP could not run to the token contract that made Remnus. - [MCP-Native vs MCP-Integrated](https://remnus.com/docs/mcp-native-vs-integrated): Two architectural approaches to AI agent access and why the distinction matters. - [Why We Chose AGPL-3.0 for Remnus](https://remnus.com/docs/why-agpl-3): Licensing philosophy: why not MIT, why not BSL, and what AGPL actually protects. ## Optional - [Pricing](https://remnus.com/pricing): Plans and limits - [Security](https://remnus.com/security): Auth methods, token scopes, and responsible disclosure - [Download](https://remnus.com/download): Desktop and mobile apps