Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者提供的说明(英文原文)
Local MCP server for requesting signatures from Agent-Q Firmware.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"agent-q": {
"command": "npx",
"args": [
"-y",
"@stelis/agent-q"
]
}
}
}在终端中运行这条命令
claude mcp add agent-q -- npx -y @stelis/agent-q粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"agent-q": {
"command": "npx",
"args": [
"-y",
"@stelis/agent-q"
]
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"agent-q": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@stelis/agent-q"
]
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日
注册名称已通过 GitHub 账户 stelis-dev 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(0.2.2)于4个月前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
同一分类中有:Spec Driven Development, Remote Desktop Commander, HotCodePush
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Manage over-the-air (OTA) updates for React Native, Expo, Capacitor and Cordova apps.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Development status: Agent-Q is an active development project with
hardware-tested Sui signing paths for CLI, MCP, and supported provider flows.
The current StackChan CoreS3 Firmware path uses DEV_PROFILE material intended
for development and demos, not real-asset custody. See Current Status for
storage and profile limitations.
Agent-Q separates execution surfaces from signing authority.
Execution capability is not signing authority. Agent-Q keeps that boundary explicit for agent-driven and local Sui signing flows.
This README describes the Agent-Q authority-boundary architecture for local Sui signing flows. Current implementation status and profile limitations are tracked in docs/IMPLEMENTATION_STATUS.md.
MCP clients, Sui CLI, and supported Sui app adapters submit bounded requests through the local agent-q host process. AI agent runtimes usually reach Agent-Q through MCP clients. The host relays requests; it does not store signing material, decide policy, or approve signing.
Agent-Q Firmware runs on a separate signing device. It holds signing material and policy for the implemented material profile, checks device state and request contents, applies the implemented policy or device-local approval gate, and returns signed results or explicit failures.
Signing request transport stays local. Signing authority stays on the device. Requesters can ask for signatures, but they cannot hold signing material, choose the authorization mode, or approve their own requests.
Agent-Q has three roles:
agent-q process. It exposes MCP stdio tools, the local HTTP API, the Admin Page, and adapter-facing request surfaces, then relays bounded requests to Firmware.Agent-Q uses a simple authority model:
For signing requests, Firmware decides whether the implemented policy or device-local approval gate permits signing.
Unsupported chains and methods fail explicitly.
agent-q-sui-signer as a Sui CLI external signer. See packages/agent-q/README.md.packages/agent-q/README.md.packages/provider-sui/README.md and packages/sample-sui-dapp-kit/README.md.packages/sample-zklogin-test-web/README.md.firmware/README.md.| Package | Use it when you want to... |
|---|---|
@stelis/agent-q-core | discover devices, open sessions, call the Agent-Q protocol, and parse Firmware results. |
@stelis/agent-q | run the local MCP server, Admin Page, and agent-q-sui-signer. |
@stelis/agent-q-provider-sui | connect a Sui app to an Agent-Q device through a provider / Wallet Standard adapter. |
packages/sample-sui-dapp-kit | run a small dapp-kit sample that signs through Agent-Q. |
packages/sample-zklogin-test-web | run a non-production browser test tool for Sui zkLogin proof preparation/proposal and sign-only transaction checks through the Agent-Q browser provider over Web Serial. |
The current implementation includes these device-owned paths. They are development/demo capabilities; Current Status below defines the DEV_PROFILE and product-active limits.
sign_transaction supports inline txBytes and same-session staged payload delivery for larger Sui transaction bytes, up to the Sui serialized transaction maximum. This is a payload and adapter input capacity, not a claim that every Sui transaction shape is semantically signable; unsupported or unbindable shapes fail before signing.flowchart LR
Agent[AI agent] --> MCP[MCP API] Admin[Admin Page] --> HTTP[HTTP API] SuiCLI[Sui CLI] --> Signer[agent-q-sui-signer] App[Sui app] --> Provider[Sui provider]
MCP --> Server[agent-q local server] HTTP --> Server Signer --> Server Provider --> Server
Server --> Core[Agent-Q core] Core --> Firmware[Firmware]
Current signing routes are summarized below. In this README, product-active means the matching source, docs, tests, build, current-tree target hardware evidence, and visual evidence are complete in docs/IMPLEMENTATION_STATUS.md. A route can be implemented and hardware-tested while product-active evidence is still being completed.
| Chain | Method | Current behavior |
|---|---|---|
sui | sign_transaction | Current implementation supports inline or same-session staged Sui transaction bytes. It validates the request, parses bounded offline transaction facts when available, applies the device-local policy or user authorization gate, and produces a signature response only after that gate authorizes signing. The parsed sender must match the active account; the parsed gas owner must also match unless the active account's device-local Sui account setting accepts gas sponsors. Policy mode fails closed when required policy coverage is missing, incomplete, unmatched, or reject-matched. User mode shows covered offline facts only when Firmware can derive the required semantic facts and fit the required review details into its bounded display summary; parser success alone is not a clear-review guarantee. Otherwise it shows a device-local blind-signing warning for a valid transaction bound to the active account. Blind signing means confirming signable bytes whose transaction details are not fully reviewable from offline facts. |
sui | sign_personal_message | Current implementation supports bounded Sui personal-message signing in user authorization mode. Policy authorization mode fails closed for this method. |
The current implementation contains Sui zkLogin active identity paths, but docs/IMPLEMENTATION_STATUS.md does not mark them product-active. The sample test web exercises browser-to-device proof setup and sign-only transaction checks over Web Serial. Enoki, OAuth, prover work, and Google OAuth testnet setup stay in the browser test layer; Agent-Q receives only bounded proof proposals. The sample does not display JWTs or proof JSON and does not write them to browser storage.
Unsupported chains and unsupported methods fail explicitly. Chains are exposed through the shared protocol; Agent-Q does not create separate chain-specific product APIs.
This section explains who may request work and who may decide. DEV_PROFILE storage and product-active status are covered in Current Status.
Detailed status lives in docs/IMPLEMENTATION_STATUS.md.
Current implementation includes the Sui signing routes listed above, MCP tools, provider-sui, and StackChan CoreS3 Firmware paths. Hardware smoke coverage is recorded in docs/IMPLEMENTATION_STATUS.md; product-active signing support is claimed only when the matching status row says the source, docs, tests, build, hardware, and visual evidence are complete.
Current limitations:
specs/PROTOCOL.mddocs/SECURITY_MODEL.mddocs/STATE_MODEL.mddocs/POLICY_SCHEMA.mddocs/IMPLEMENTATION_STATUS.mdfirmware/README.mdfirmware/src/stackchan-cores3/README.mdInstall workspace dependencies:
npm install
Build all packages:
npm run build
Run the root test suite:
npm test
Firmware build instructions are target-specific. Start with firmware/README.md and the target README under firmware/src/ /.
.WORK/ is for local planning, scratch files, and investigation materials. It is not tracked by Git and must not be required by user-facing build or runtime instructions.