Devotd Github
NewAI agents

withnative/native — a workspace and MCP server built on its own context

withnative/native is an open-source workspace engine and MCP server that connects the documents a user already has with the model of their choice.

Repositorywithnative/native ↗

withnative/native is an open-source workspace engine and MCP server that connects the documents a user already has with the model of their choice.

GitHub description: ChatGPT Space, but open source and more powerful. Own your context, bring any model and shape any UI. Engine + MCP server, with desktop app coming soon..

stars
★ 74
Forks
0
Language
Rust
License
NOASSERTION
Read original source ↗View on GitHub ↗

withnative/native is an open-source project that provides a document-based workspace that users either hold themselves or self-host. The current repository contains an engine and MCP server that connect a specific model with custom screens, while the desktop app is described as being in preview. The repository's primary language is Rust. No latest release is listed in FACTS.

According to the README, this workspace is designed to connect tasks, projects, and messages to one another, and to produce real-time data documents and dashboards, collaborative editing canvases, and custom screens. Its self-hosted operation stores data in a local SQLite file and connects with MCP clients. The developers state that in their own search benchmark comparing against a Markdown-based store, token usage per correct answer was 40 times lower. That is a configuration worth watching for users who want to manage context directly and connect multiple models and screens.

That said, the 40x figure comes from the developer's own published benchmark, and FACTS gives no measurement conditions or comparison details. The license is also recorded as NOASSERTION, so reuse terms cannot be confirmed. The desktop app is still in preview. The scope that can be confirmed today is therefore centered on the engine and the MCP server, and it is difficult to accept either the benchmark result or a finished desktop app as verified performance and features.