Native

MCP文档知识

提供方 未标注

文档知识 未标注 未标注 社区 / 第三方 ClaudeCursorVS CodeCodex

面向智能体的工作区引擎:文档、决策、任务和消息以类型化、相互关联的记录形式存储在由你掌控的本地 SQLite 文件中,保留完整变更历史,并通过 MCP 执行 SQL 查询。

一句话选型结论

按资料字段整理

Native 属于文档知识类 MCP 服务器,传输方式未标注,鉴权方式未标注,可直接接入 Claude、Cursor、VS Code 等,没有现成安装命令,需按官方文档接入。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类文档知识 传输方式未标注 鉴权未标注 部署方式未标注
GitHub 星标 未标注 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

文档知识
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

支持客户端

适用场景

记忆

安装与配置

未标注 未标注
上游未给出传输方式
  1. 1
    准备运行环境

    按官方文档确认运行环境(常见为 Node.js 或 Python),装好后再执行安装命令。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
文档检索问答

在你的文档库里按问题找答案。

知识库读写

把新资料写进知识库,随用随取。

内容整理

把零散资料归纳成结构化笔记。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 已适配 Claude、Cursor、VS Code,换客户端不用重新研究配置。

先想清楚的地方

  • 没有现成安装命令,需要按官方文档自己填启动命令和参数。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在文档知识场景里让模型直接调用外部能力的人。
  • 已经在用 Claude、Cursor 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Native当前 未标注 未标注 未标注 无 2026-11-20
dify 未标注 stdio · 本地进程 本地部署 有 2026-04-11
open-webui 未标注 stdio · 本地进程 本地部署 有 2026-04-11
JeecgBoot 未标注 stdio · 本地进程 本地部署 有 2026-04-11
Cherry Studio 未标注 未标注 未标注 无 2026-05-17
mempalace 未标注 stdio · 本地进程 本地部署 有 2026-04-11

服务器档案

文档知识
字段来自项目仓库与公开索引,中文说明由本站标注
提供方未标注上游标注的维护方 未标注
分类文档知识本站按上游分类展示
连接方式未标注上游未给出传输方式
鉴权方式未标注远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · 未标注本地安装时用到的包标识与运行时
部署类型未标注上游给出的部署形态
最近核验2026-11-20上游最后核验时间

详细介绍

8.0 KB
来自项目仓库与上游目录的原文,未做删改

Native

Open-source ChatGPT Space. Powered by your Markdown vault.

Join the waitlist → ·
Watch the demo

Why we built Native

ChatGPT Space is an
AI-native
workspace from OpenAI.

We think it just needs three changes:

  Problem
  Solution

  Context locked to one provider
  Own and self-host your context

  Limited to OpenAI models
  Bring any model

  Another interface to switch to
  Choose your interface

See our research:
40x more token-efficient than Markdown

Docs ·
Contributing ·

Longer stuff we haven't made more concise yet

ChatGPT Space, made open-source and more powerful

An agent-native workspace for documents, projects and messaging.

Work and build shared context together, in one place. Every teammate, any AI agent.

> 40× fewer tokens than Markdown vaults
>
> In our context retrieval benchmark, agents using Native consumed 40× fewer tokens per fully correct answer than agents searching a Markdown vault with wikilinks.

Ask ChatGPT about Native

An open-source alternative to ChatGPT Space, with extra power.

  • Live, self-updating documents, slides and dashboards that read and visualise data live from your workspace
  • Easily connected context that brings together your tasks, projects and team messaging
  • Granular sharing options for your team and their agents to view and edit
  • UIs that you can mould however you like: e.g. as Google Docs, Linear, Notion or Slack
  • Use your existing subscriptions and work with ChatGPT, Claude, Copilot, Muse and many more

Rich context, in any form you like

In Native, context isn't a pile of files for an AI to read. It's live, connected data that you, your teammates and your agents can see in whatever form suits the work:

  • Documents with live data in them. A status update can show the current state of the tasks it talks about, rather than a copy that goes stale.
  • Slides, dashboards and other artifacts written in HTML or MDX, which read real records from your workspace as they render. When the work changes, the deck changes. (artifact runtimes)
  • Canvases where notes, shapes and live cards for real records sit side by side, edited by people and agents at the same time. Experimental. (canvas protocol)
  • Interfaces you invent. The same tasks can be a Kanban board, a roadmap or a bookshelf at once, without copying them into another tool. Ask your agent to build the view you want.

Artifacts only see the records they're given and the person viewing them is allowed to read. Any edit they make goes through the same permission checks and history as everything else.

Preview access

Native is in preview access.

This repo has our open-source MCP server, and it will soon have our desktop app (shown in the demo video).

Currently, desktop app access is limited to a select number of preview testers.

You can join the waitlist to get notified about the desktop app's release. You may also be invited to be a preview tester.

Try the agent experience today

You don't need to wait to see how your agents will work with Native. This repository has the Native engine and its MCP server, which you can run on your own machine against a SQLite file you own. It's a preview of the agent experience that will come built into the desktop app.

Give your coding agent this:

> Help me run Native locally from https://github.com/withnative/native.
> Follow SELF_HOSTING.md. Build mcp-stdio, keep its SQLite database in a
> directory I choose, connect it to this client over stdio, and prove it works
> by creating a record, restarting, and reading it back.

Then ask it to record a decision, open a task that depends on it, show you the history of both, and build an artifact that reads them.

Already have access?

Install the Native plugin in Claude Code or Codex:sh claude plugin marketplace add withnative/plugins && claude plugin install native@withnative codex plugin marketplace add withnative/plugins && codex plugin add native@withnative

Or paste this into Claude or ChatGPT/Codex: Open https://github.com/withnative/native-plugin and follow the setup guide for the Native plugin. The setup guide covers sign-in and other clients.

A Native way of working

  • Keep everything together. Your documents, projects and team messaging are connected, both in your UI and your agent's tool calls.
  • Work however you're used to. Write directly in the UI yourself, talk to ChatGPT / Claude, or orchestrate agents via the terminal.
  • Cascade context in realtime. Shared changes are visible to your teammates and their agents, automatically. No sync step.
  • Trace and restore any change. Audit the work done by AI and roll it back effortlessly if it's gone wrong.
  • Mould your tools to your work. Use custom UI and plugins written for you by your AI, with no redeployment work.

How it works under the hood

  • A single, queryable database. Your documents, projects and messages are in one, connected place. (src/query, src/store.rs)
  • Queries, not browser screenshots. Your agents interact with Native through SQL and MCP. (src/mcp/tools, query guide)
  • Typed, structured graph. Your context is connected via directed relationships, like supersedes and depends_on. (src/relationship, src/schema)
  • Per-node visibility grants. Choose which records to share with your team, and who gets to see or edit them. (src/authorization.rs, crates/policy-kernel)
  • Event log as the source of truth. Everything is tracked, with full replayability and transparent trails. (src/events.rs, src/projector, temporal reads)
  • UI as an arbitrary view of the data. We use default renderers for messaging, document and project views, but you can write your own! (artifact runtimes)

For the full map, see ARCHITECTURE.md and the capability map, which pairs each claim with its code, tests and limits.

What you can run today

You want to Today
Run the engine on your own machine, with your own MCP client, against a SQLite file you own Works from this repo. See Try the agent experience today.
Use Native with Claude, ChatGPT or another MCP client on a hosted workspace Preview testers only. Join the waitlist to be notified, or invited to test.
Use the desktop app from the video Preview testers only. Coming to this repo soon; join the waitlist to hear when it's released.
Self-host the full product: apps, sign-in and teams on your own server Not yet. See the roadmap.

Public source snapshot. This repository holds the Native engine: the SQLite reference node, the local MCP server, the full tool implementation, the event log and projections, search and queries, export, and the federation protocol work. The hosted service, accounts and sign-in, the workspace apps and the desktop app are not in this snapshot yet; only the optional experimental MCP App bundles are included. Every published file is listed in native-boundary.json.

Roadmap

  • The desktop app and default apps, open-sourced here soon.
  • Self-hosting the whole product, including the apps, sign-in and team membership on your own server, not only the engine.
  • More composable. Plugins for both the apps and the engine.
  • Workspaces that talk to each other. Separate Native servers that exchange messages while each keeps its own data. See federation transport.
  • A paid managed cloud for teams who'd rather not run infrastructure, always with an easy export so you can move to self-hosting.

Licence

Native's source is licensed under the GNU Affero General Public License v3.0 only (AGPL-3.0-only); see LICENSE.md.

Contributing and contact

Native is developed in private repositories and published here as read-only snapshots, without their development history.

Contributions are by invitation; see the contribution policy and the snapshot notes. You don't need to contribute here to build your own apps or plugins.

Issues and pull requests are not the place for support. Email richard@withnative.ai or visit withnative.ai.

常见问题

6 问
答案由本站依据条目字段整理,不引用上游原文
Native 怎么安装?

资料没有给出安装命令,请到官方仓库或文档按说明接入;这类条目多数是生态应用,或尚未发布可安装包。

Native 怎么在 Claude、Cursor、VS Code 里配置?

它的传输方式未标注,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

Native 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Native 需要 API Key 吗?

资料没有标注鉴权方式,是否需要密钥取决于服务本身,接入前按官方文档确认。

Native 支持哪些客户端?

资料标注支持 Claude、Cursor、VS Code、Codex、Zed;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

Native 能提供哪些能力?

资料显示能力标签包括 记忆。具体工具清单以官方文档为准。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-11-20。