Sutra

io.github.shivnathtathe由 GitHub 账号 shivnathtathe 发布v1.0.1更新于 Oct 24, 2026

Private, portable memory and reusable skills for AI agents.

已验证Streamable HTTP可网页运行知识与记忆AI 与机器学习

概览

AI 生成的概览

托管式 Sutra 服务，通过 10 个带类型的 MCP 工具为 AI 助手提供私密、可迁移的记忆与可复用技能。

可以做什么
  • 读取共享身份上下文、按关键词搜索记忆，并列出或加载指定的项目集群
  • 把持久的偏好、决策、纠正与事实保存为观察记录
  • 导出全部已存记忆，便于数据迁移
  • 发现用户已启用的技能，加载技能及其资源索引，并读取单个技能资源
  • 在明确授权后上传新的不可变技能版本
适用场景
适合希望上下文跨会话、跨不同 AI 客户端延续的场景，也适合维护多个助手共用的可复用技能。适合想要厂商中立、可导出记忆库而非各客户端独立历史的用户。
运行要求
需要远程 Streamable HTTP 端点，无需本地安装包。需要 Sutra 账号。认证方式为 OAuth 2.1（浏览器授权）或按客户端使用放在 Authorization 头中的专用可撤销 API 密钥。需能访问该托管服务。
安装前请注意
服务会保存助手写入的内容，请留意实际保存了什么；写入、上传与导出操作本应需要用户明确意图。不要把 API 密钥或 OAuth 令牌写进纳入版本控制的配置文件，应使用客户端的密钥保管机制。导出会生成全部已存记忆的完整副本。接入敏感工作流前请阅读安全、隐私与服务条款文档。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Sutra，将其添加到工作区。
  2. 连接 所需的 凭据。它们会被私密存储，绝不会在公开页面上展示。
  3. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.shivnathtathe%2Fsutra/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。 客户端会在首次连接时完成 OAuth 登录。

{
  "mcpServers": {
    "sutra": {
      "type": "http",
      "url": "https://sutra-c3or.onrender.com/mcp"
    }
  }
}

README

[Sutra - one private memory layer for every AI agent]

Open MCP integration files for the hosted Sutra service.
Connect private, portable context to the AI agents you already use.

Website  ·  All integrations  ·  Smithery  ·  Glama  ·  Support

[MCP Streamable HTTP] [OAuth 2.1] [Ten typed tools] [MIT licensed integrations]

Connect Sutra

text
https://sutra-c3or.onrender.com/mcp
ConnectionUse whenWhat happens
OAuth 2.1Your client supports remote MCP authorization.The client opens Sutra in your browser, requests consent, and stores its grant.
API keyYour client expects an Authorization header.You create a dedicated, revocable key in Sutra and store it with the client's secret mechanism.
  1. Create a Sutra account.
  2. Choose a client configuration below.
  3. Complete OAuth or enter a dedicated API key when prompted.
  4. Confirm the client discovers 10 tools.
  5. Try: Use Sutra to list my project clusters. Ask before saving anything.

[!IMPORTANT] Never put a real Sutra API key, OAuth token, memory export, or password in a checked-in configuration file.

Choose Your Client

ClientAuthenticationConfiguration
CursorOAuthclients/cursor-mcp.json
OpenCodeOAuth with DCR and PKCEclients/opencode.json
VS Code / GitHub CopilotRevocable API keyclients/vscode-mcp.json
Claude CodeRevocable API keyGuided CLI setup
Cursor   OAuth

Install this repository as an Agent Plugin, or add the following to ~/.cursor/mcp.json:

json
{  "mcpServers": {    "sutra": {      "url": "https://sutra-c3or.onrender.com/mcp"    }  }}

Connect the server and complete Sutra's browser authorization flow.

OpenCode   OAuth + DCR + PKCE

Merge clients/opencode.json into your global or project opencode.json:

json
{  "$schema": "https://opencode.ai/config.json",  "mcp": {    "sutra": {      "type": "remote",      "url": "https://sutra-c3or.onrender.com/mcp",      "oauth": {}    }  }}

Authorize and inspect the connection:

sh
opencode mcp auth sutraopencode mcp list
VS Code and GitHub Copilot   API key

Copy clients/vscode-mcp.json to .vscode/mcp.json, or merge it into your user MCP configuration:

json
{  "inputs": [    {      "id": "sutra-api-key",      "type": "promptString",      "description": "Sutra API key",      "password": true    }  ],  "servers": {    "sutra": {      "type": "http",      "url": "https://sutra-c3or.onrender.com/mcp",      "headers": {        "Authorization": "Bearer ${input:sutra-api-key}"      }    }  }}

Start Sutra from the MCP configuration editor. VS Code requests the key through a masked input instead of storing it in the file.

Claude Code   API key

Use the Sutra CLI for guided setup:

sh
npm install -g @shivnathtathe/sutra-clisutra connect claude-code

The CLI requests a dedicated key and safely writes the supported local configuration. It does not send your key to this repository.

Need another agent? Sutra publishes current instructions for all 22 tested integrations.

Confirmed Compatibility

[The 22 AI clients independently tested with Sutra]

11 web agents use OAuth. 11 CLI and desktop agents use dedicated API keys. Product names and marks belong to their respective owners.

Tool Surface

Sutra exposes 10 typed tools. Read operations retrieve only the context needed for the current request. Write, upload, and export operations require deliberate user intent.

Context and memoryAccessPurpose
read_contextReadLoad the user's shared identity context.
search_memoryReadFind targeted memories by keyword.
list_clustersReadDiscover available project clusters.
get_clusterReadLoad one known project cluster.
write_observationWriteSave one durable preference, decision, correction, or fact.
export_allExportExport all user memories for data portability.
Reusable skillsAccessPurpose
list_skillsReadDiscover skills explicitly enabled by the user.
get_skillReadLoad an enabled skill and its resource index.
get_skill_resourceReadRead one resource referenced by an enabled skill.
upload_skillWriteUpload a new immutable skill version after explicit authorization.

Data Boundaries

Sutra doesSutra does not
Stores observations deliberately saved by a user or authorized agent.Copy every conversation from a connected client.
Isolates unrelated context in named project clusters.Share one cluster with every request automatically.
Shows the source agent, tags, and activity for stored memories.Hide writes inside an opaque client profile.
Supports revocable OAuth grants and per-client API keys.Require one permanent credential across all agents.
Provides an explicit full export for portability.Lock user memory to one AI vendor.

Read Security and data boundaries, the Privacy Policy, and the Terms of Service before connecting sensitive workflows.

Repository Contents

text
.|-- plugin.json             # Agent Plugins 1.0 manifest|-- mcp.json                # Portable Streamable HTTP connection|-- server.json             # Official MCP Registry metadata|-- clients/|   |-- cursor-mcp.json|   |-- opencode.json|   `-- vscode-mcp.json|-- docs/|   |-- client-support.md|   `-- security-and-data.md`-- assets/    |-- integration-grid.svg    |-- readme-hero.png       # Pixel-tested GitHub banner    |-- readme-hero.svg       # Editable banner source    `-- sutra-logo.png

Open Integration Layer

This repository contains the open-source manifests, examples, documentation, and artwork needed to connect clients to Sutra. Corrections for changing client formats are welcome; see CONTRIBUTING.md and SECURITY.md.

The files in this repository are licensed under the MIT License. The hosted Sutra backend, dashboard, APIs, infrastructure, datasets, service implementation, and trademarks remain proprietary and are not included. The Sutra logo and third-party product names and marks are excluded from the MIT grant; see NOTICE.

Context should outlast the chat window.
Create your private workspace

来源：README.md，提交 f706b30

工具

0
该服务器需要登录，因此 SourceWeft 无法在没有账号的情况下读取其工具，也未收录。已连接账号所能看到的内容可能因账号及其权限而异。

版本历史

1
  1. v1.0.1最新Sep 16, 2026

评分

已满足适用检查项的 85%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list，或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分