gno.sh

What It Is

GNO is a local knowledge workspace that indexes Markdown, PDF, Word, Excel, PowerPoint, plain-text files, and code directories into a searchable knowledge base. It emphasizes “paragraph-level” retrieval rather than simply finding files, and can use a local LLM to generate answers with citations. Local GNO itself is open source, MIT-licensed, and free forever, while gno.sh is a separate hosted publishing layer.

Core Capabilities and Integrations

Its core strength is hybrid retrieval: a combination of keyword search, semantic vectors, and reranking, with support for search, vsearch, query, and structured queries. AI features run locally via node-llama-cpp, with built-in model presets such as slim-tuned, slim, balanced, and quality. It also supports Hugging Face GGUF, local GGUF files, and OpenAI-compatible HTTP backends such as Ollama, LM Studio, and vLLM. Integrations are fairly comprehensive, including CLI, Web UI, REST API, SDK, MCP, and agent skills. It can connect to Claude Code, Claude Desktop, Claude Cowork, Cursor, Codex, Zed, Windsurf, and more, allowing AI tools to retrieve personal documents on demand—similar to a persistent memory layer.

Pricing and Privacy

Local GNO is fully featured at $0, with no account, no API key, and no telemetry. Indexing, embeddings, search, and Q&A all run on the local machine and can be used offline. The free tier of gno.sh publishing includes 10 notes, 3 collections, and 1GB of storage. Pro costs $9/month and unlocks secret links, invite-only spaces, 50GB of storage, and more. Team costs $29/month and includes end-to-end encrypted sharing, white labeling, custom domains, team seats, and SLA support. There is currently no trial for Pro or Team.

Pros, Cons, and Who It’s For

Its advantages include clear privacy boundaries, open-source and free local use, a professional retrieval pipeline, citation-based answers that help with traceability, and deep integration with mainstream AI coding and chat tools. The limitations are that installing and running local models has a learning curve, results depend on hardware and model choice, Office/PDF source files cannot be edited in place, Windows arm64 is not currently supported, and the Chinese UI and Chinese-language performance are not clearly documented. It is especially suitable for researchers, consultants, developers, Obsidian users, LLM Wiki users, and private RAG scenarios where users do not want to upload materials to the cloud.

Access in China

The source text does not provide details on network availability or payment methods in mainland China, so its access status is rated as unknown. Since Local GNO can run offline, core usage does not depend on the cloud. However, model downloads, GitHub/npm, Hugging Face, or the gno.sh publishing service may be affected by network conditions in China. If access is restricted, alternatives include Obsidian, local Ollama + RAG setups, or domestic knowledge-base tools.