Local Knowledge Engine

Your Notes, Docs, Agents, and Search Stack in One Local Workspace

Search, browse, graph, edit, capture, and cite across everything you own.

GNO has grown beyond a hybrid search CLI. It is now a full local knowledge workspace with a fast web UI, app-level tabs, cross-collection browse tree, knowledge graph, AI answers, and direct integrations for coding agents and MCP clients.

100% local + private

Search + Graph real daily-driver workspace

Skills + MCP agent memory without cloud drift

GNO web UI dashboard with search, tabs, and workspace surfaces

Current workspace Search, tabs, capture, and health in one shell

GNO cross-collection browse workspace tree and collections view

Browse Cross-collection tree + folder detail panes

Graph Wiki, markdown, and similarity edges

Get Started View on GitHub Explore the Workspace

Works with your tools

Claude Code Codex OpenCode OpenClaw Cursor Claude Desktop Zed Windsurf Amp Raycast

For humans

Stop bouncing between Obsidian, Finder, grep, and browser tabs

Search, browse, graph, edit, and capture in one local product surface.

For agents

Give Claude Code, Codex, Cursor, and MCP clients a real memory

Skills for low-overhead retrieval, MCP for richer tool access, both on your own files.

For builders

One index, many surfaces

CLI, REST API, SDK, web workspace, desktop shell, and agent integrations all point at the same local knowledge engine.

Three Commands to Searchable Knowledge

Point

Add any directory—Obsidian vault, project docs, research papers, code repos.

gno init ~/vault --name notes

Index

GNO builds BM25 + vector indexes. Understands 30+ languages. Incremental re-indexing.

gno index

Search

Hybrid search from CLI, Web UI, REST API, or any AI agent via MCP / Skills.

gno query "auth best practices"

Built for How You Actually Work

For Knowledge Workers

Your Obsidian Vault, Supercharged

You have thousands of notes across Obsidian, Logseq, or plain Markdown folders. GNO indexes them all and adds what your note app can't—hybrid semantic search, AI-powered answers with citations, and a knowledge graph that reveals connections you didn't know existed.

gno query "what did we decide about the API redesign" gno ask "summarize my meeting notes from last week" --answer

How search works →

For AI Agents

Give Your Coding Agent a Memory

Claude Code, Codex, OpenCode, OpenClaw, Cursor—they're powerful but amnesiac. Install GNO as a skill or MCP server and your agent can search your entire knowledge base on demand. Zero context pollution, instant retrieval.

gno skill install --scope user gno mcp install --target cursor

Agent integration →

For Developers

Embed Search in Your App

Use GNO as a TypeScript SDK, REST API, or CLI pipe. Build internal tools, automate knowledge retrieval, or add semantic search to any workflow. JSON output everywhere, scriptable and composable.

gno search "JWT token" --json | jq '.results[].uri' curl localhost:3000/api/query -d '{"q":"auth"}'

SDK docs →

What GNO Feels Like Now

Daily-Driver Workspace

App-level tabs, quick switcher, safe editing, capture flows, and a browse tree that finally makes the workspace feel spatial.

Workspace tour →

Real Browse + Graph Surfaces

Search is no longer the only entry point. Browse collections as a tree, inspect folders directly, and jump through a knowledge graph.

See graph view →

Agent-Ready by Default

Skill installs for coding agents, MCP server for desktop/editor clients, and retrieval that stays grounded in your own local corpus.

Agent integration →

Builder Surface Included

SDK, API, CLI, and desktop/web surfaces stay aligned, so internal tools and local products can reuse the same index and retrieval stack.

SDK + API →

Everything You Need

Hybrid Search

BM25 full-text + vector similarity with reciprocal rank fusion and cross-encoder reranking.

AI Answers

Cited, grounded responses from your own documents. Local LLM or remote GPU—no cloud APIs.

Privacy First

100% local. Zero telemetry. Works offline. Your documents never leave your machine.

Agent Skills

One-line install for Claude Code, Codex, OpenCode, OpenClaw. Zero context pollution.

MCP Server

19 tools for Claude Desktop, Cursor, Zed, Windsurf, Amp, Raycast, and more.

Multi-Format

Markdown, PDF, DOCX, XLSX, PPTX, plain text. Frontmatter extraction and language detection.

Web UI

Dashboard for search, browse, edit, and AI Q&A. Knowledge graph visualization. Keyboard-first.

Fine-Tuned Models

Published retrieval model on Hugging Face. Train your own with local MLX LoRA.

Benchmarks

Measure retrieval and embedding changes against fixed corpora, real GNO code, and public OSS slices before changing defaults.

Knowledge Graph

Interactive visualization of wiki links, markdown links, and semantic similarity edges.

Tags & Filtering

Hierarchical tags from frontmatter. Temporal, author, and category filters. AND/OR logic.

REST API & SDK

35+ HTTP endpoints. TypeScript SDK for embedding GNO in your own apps. JSON everywhere.

Fast CLI

Bun-powered instant startup. Search 15,000+ docs in milliseconds. JSON, CSV, Markdown output.

Quick Start

# Install
bun install -g @gmickel/gno

# macOS: Vector search requires Homebrew SQLite
brew install sqlite3

# Point at your files
gno init ~/obsidian-vault --name vault
gno init ~/projects/docs --name docs

# Index everything
gno index

# Search
gno query "authentication best practices"
gno ask "summarize the API discussion" --answer

# Give your AI agent access
gno skill install --scope user          # Claude Code / Codex / OpenCode
gno mcp install --target cursor         # Cursor / Zed / Windsurf

Read the Full Documentation →

One Index, Four Real Surfaces

Hybrid retrieval that actually scales

BM25, vector similarity, query expansion, reranking, structured query controls, and grounded answers from one local index.

Workspace

Browse, edit, capture, and graph in the same app surface

Recent tabs, collection tree, safe editing, read-only converted docs, frontmatter-aware metadata, and a knowledge graph that feels connected to the rest of the product.

Agents

Skills for low-overhead retrieval, MCP when you want tools

Claude Code, Codex, OpenCode, OpenClaw, Cursor, Claude Desktop, Zed, Windsurf, Amp, Raycast, and more.

Builder

CLI, API, SDK, and desktop/web all point at the same engine

Use it from the shell, embed it in Bun apps, wire it into internal tools, or keep it running headlessly with daemon mode so agents and automations always see a fresh local index.

Browse

Cross-collection tree browsing

The web workspace now behaves like a real note workspace, not a flat result list. Move through collections and nested folders, keep separate browse context per tab, and jump from folder structure straight into the note you want.

Cross-collection tree sidebar
Folder detail panes
Pinned collections and favorites
Deep-linkable browse state

GNO browse workspace with cross-collection tree

Graph

Connections, not just matches

See wiki links, markdown links, and semantic similarity in one navigable constellation. The graph is part of the same workspace, so you can move from node discovery straight into the source note.

Interactive node navigation
Collection filtering
Optional similarity edges
Document-level jump targets

Why Teams Stick With It

Local by default

No cloud indexing, no hosted memory layer, no telemetry requirement. Your files stay on your machine.

Safe editing model

Markdown and plaintext stay editable, converted binaries stay read-only, and the UI makes the distinction clear.

One retrieval stack everywhere

CLI, SDK, web UI, desktop shell, and agent integrations all share the same index and retrieval behavior.

Good for humans and agents

Use it as a personal workspace, or let your coding agent query the same knowledge base without bloating its prompt.