nexwiki

NexWiki

The personal context consolidation repository and AI verification foundation.

---

🧠 Why NexWiki?

Personal, individual context is scattered across disconnected tools—random scratch notes in note applications, loose Markdown files strewn across drives and folders, technical decisions buried in Microsoft Teams or Slack channels, and critical snippets inside email messages.

NexWiki's mission is to give you a single, local, human-readable place to relocate and unify all of that data—even when raw, unformatted, or incomplete.

Connect your AI agent harness apps—including Claude Code, GitHub Copilot CLI, Google Antigravity, OpenAI Codex, OpenCode, and Cursor—to ingest, organize, and reason over this unified context. This establishes a powerful dual payoff:

1. Accelerate Building: Leverage your accumulated knowledge and personal context to help you design and build new solutions faster.
2. Ground Truth Verification: Empower your agents to use that personal context to verify that what they are building is actually correct and compliant with your standards.

🖥️ Interactive Web UI

Experience creating, formatting, and organizing articles in NexWiki's responsive web interface.

🤖 AI Agent MCP Ingestion

Watch an AI agent harness ingest an unformatted Slack decision snippet into NexWiki over MCP and instantly query the newly grounded context.

---

⚡ Quick Start

Get NexWiki up and running on http://localhost:5808 in seconds.

1. Installation

Binary Installation (Recommended)

Installs the pre-compiled, SHA256-verified standalone binary for your OS and architecture:

• macOS / Linux:

  curl -fsSL https://raw.githubusercontent.com/titusai-io/nexwiki/main/scripts/install.sh | bash
  

  (or run ./scripts/install.sh from a cloned repo)

• Windows (PowerShell):

  irm https://raw.githubusercontent.com/titusai-io/nexwiki/main/scripts/install.ps1 | iex
  

  (or run .\scripts\install.ps1 from a cloned repo)

Docker Container Runner (Alternative)

Pulls the latest multi-arch image (ghcr.io/titusai-io/nexwiki:latest) and automatically mounts your OS data directory:

• macOS / Linux:

  curl -fsSL https://raw.githubusercontent.com/titusai-io/nexwiki/main/scripts/docker-run.sh | bash
  

  (or run ./scripts/docker-run.sh from a cloned repo)

• Windows (PowerShell):

  irm https://raw.githubusercontent.com/titusai-io/nexwiki/main/scripts/docker-run.ps1 | iex
  

  (or run .\scripts\docker-run.ps1 from a cloned repo)

2. Launching

Start the server and automatically open the application in your default web browser:

nexwiki -launch-in-browser

On first launch, NexWiki automatically creates its data directory (articles/, assets/, history/, and search index) in your OS standard directory (~/.config/nexwiki/nexwiki-data on macOS/Linux, %AppData%\nexwiki\nexwiki-data on Windows, or /app/data in Docker) and seeds an initial homepage. The seed only happens when the article directory is empty, and a document is only served at the path its own slug dictates — a home.md whose front matter declares a different slug is not served at the home path (the dashboard 404s) and is never reseeded, because a document is only valid at articles/<slug>.md.

📘 For advanced CLI flags, environment variables, manual downloads, and production Docker Compose setups, see the Install Scripts Guide, Configuration Guide, and Docker Deployment Guide.

---

🤖 Connecting AI Agents via MCP

NexWiki contains an always-on Model Context Protocol (MCP) server exposing 7 built-in tools for search, content retrieval, memory management, and agent planning.

It implements the current 2026-07-28 MCP revision and the older initialize-based revisions on the same endpoint, so new and old clients work side by side with no configuration. Alongside tools it serves resources you can @-mention, prompts, live subscriptions over both transports, cursor pagination, and argument completion — type a few characters and your client completes the wiki page you meant. See docs/mcp_server.md.

Streamable HTTP (Recommended)

Reuses your running web server at http://localhost:5808/api/mcp and sidesteps search index locks:

{
  "mcpServers": {
    "nexwiki": {
      "url": "http://localhost:5808/api/mcp"
    }
  }
}

Add directly in Claude Code:

claude mcp add --transport http nexwiki http://localhost:5808/api/mcp

Stdio Subprocess

To run a stdio MCP server alongside your running web instance — or standalone, opening the data directory itself when no web server is found:

{
  "mcpServers": {
    "nexwiki": {
      "command": "nexwiki",
      "args": ["-mcp-only", "-data", "/path/to/data"]
    }
  }
}

(Note: Passing -mcp-only is required for stdio subprocesses. It skips the web port bind, and if a NexWiki web server is already running, it proxies to it instead of opening the data directory; see Sidecar proxy mode).

Universal Agent Skill & Operator Conventions

• Agent Skill: Copy agent-skill/nexwiki/ into your agent's skills folder (~/.claude/skills/, ~/.copilot/skills/, .agents/skills/, etc.). Built on the open Agent Skills standard, it works across Claude Code, GitHub Copilot CLI, Google Antigravity, OpenCode, and OpenAI Codex. See the Agent Skill Guide.
• Operator Conventions as Memories: Your own rules for agents — house style, corrections, how you like work done — are ordinary feedback and user memories. get_wiki_overview, the first call every agent is told to make, returns them as pinned_memories, and agents follow them. Edit one in your browser and every connected agent sees it on its next session, with no restart. The universal rules (what a memory must carry, that content replaces the whole body, what a [[WikiLink]] may point at) are built into the tools. See the AI Agent Integration Guide.

---

✨ Key Features

• 📦 Zero-Dependency Go Binary: Standalone Go executable with the React 19 SPA embedded via go:embed—no external asset servers or database processes required. (Developer Guide)
• 🤖 9 Built-in MCP Tools: Comprehensive semantic tools over Streamable HTTP and stdio, providing agents safe search, structured reads, optimistic writes, and rollback controls. (MCP Server Guide)
• 📂 Flat-File Markdown & OKF v0.2: All pages persist as human-readable Markdown files with conformant Open Knowledge Format front matter and five trust signals (provenance, trust, freshness, lifecycle, and computations). (OKF Guide)
• 🧠 Isolated AI Agent Memories & Custom Skills: Dedicated storage for agent plans, skills, and memories, cleanly segregated from standard notes with two-axis memory classification (memory_kind and memory_type). (Skills Guide · Tags & Memory Guide)
• 🕒 Gzipped History & Visual Diffs: Compressed .md.gz revision snapshots with split or unified visual diffs, instant rollbacks, and optimistic locking conflict guards. (Version Control Guide)
• 🧹 History Redaction & Audit: Purge a document's earlier revisions without deleting it (save_article(purge_history: true)), and prove a term is gone from every stored revision with search_wiki(include_history: true). (History Redaction Guide)
• 🔍 Bleve Full-Text Search Engine: Embedded Bleve index with real-time indexing, boolean query expressions, term scoring, and highlighted matching snippets. (User Guide)
• 📡 Real-Time SSE Updates & Activity Drawer: Single-connection Server-Sent Events push live document edits, glowing unread badges, and detailed audit history to the slide-in Activity Drawer. (Editor & Activity Guide)
• 📊 Responsive 4K Reader & Mermaid Diagrams: Viewport-scaling reading column for ultra-wide displays, native theme-aware Mermaid SVG diagram rendering, and clean PDF exports. (Reader & Dashboard Guide)
• 📋 Rich Text Clipboard & Formatted Sharing: One-click "Copy as Rich Text" in the Share & Export menu writes both HTML and plain text to the clipboard for formatted pasting directly into rich-text applications like Microsoft Teams, Outlook, Word, and Slack without raw markdown syntax. (Rich Text Sharing Guide)

---

📚 Documentation Hub

Explore in-depth technical manuals, architectural specifications, and workflow guides in our Documentation Hub:

• User & Content Creation Guide — Managing articles, trust signals, backlinks, and export options.
• Second Brain Workflow Guide — Orientation loop, progressive disclosure, and source ingestion.
• Model Context Protocol (MCP) Reference — Canonical reference for all 7 tools, annotations, and schemas.
• Agent Integration & SOP Guide — Connect-time instructions, operator memories, and agent blueprints.
• Plan Lifecycle & State Machine Guide — Enforced plan statuses, auto-archiving, and retention timers.
• Open Knowledge Format (OKF v0.2) Guide — Trust tiers, freshness expiration, and bundle import/export.
• Rich Text Sharing & Clipboard Guide — Formatting interoperability, clipboard APIs, and sharing to Teams/Slack/Outlook.
• Configuration & CLI Flags Reference — Complete guide to environment variables, flags, and storage defaults.
• Docker Deployment Guide — Container setup, multi-arch images, and volume persistence.
• Production Deployment & Reverse Proxy Guide — Caddy/Nginx reverse proxy, TLS, and auth boundaries.
• Developer Setup & Build Automation Guide — Local development, Vite hot-reloading, and cross-compilation.
• Release Engineering & Versioning Guide — CI quality gates, tag automation, and semantic versioning.
• History Redaction & Audit Guide — Purging old revisions, auditing every revision for a term, and fixing a document's type in place.

---

📄 License & Contributing

NexWiki is released under the MIT License and maintained by Titus AI LLC. Contributions are welcome: read CONTRIBUTING.md for what fits NexWiki's scope, and sign the Contributor License Agreement with your first pull request.