World Monitor
Real-time global intelligence on conflict, markets, energy, trade, shipping and cyber. 100 tools.
Read-only checks of an AI agent's Sigvara standing, score, stake and slashes before you rely on it.
列表按使用分(满分 100)排序,由以下部分组成
| 热度 | 20/80 | GitHub 星标或每周下载量:较大的计 60,较小的计 20 |
|---|---|---|
| 新鲜度 | 12/12 | 30 天内有发布或代码更新得满分 |
| 仍在响应 | 0/8 | 上次检查时云端端点有响应 |
| 使用分 | 32/100 |
质量: 基础检查通过 3/3 项(仅展示,不用于排序) · 上升势头: 暂无数据 (只用于“上升中”标签)
选择你使用的 AI 应用,然后复制设置。
用一句话让 AI 帮你安装
请帮我安装 Sigvara。按照 https://mcp.popnx.com/i/runtimeadmin-sigvara.md 操作,先把命令给我看,等我同意后再运行。粘贴到能在你电脑上工作的 AI(如 Claude Code、Cursor)。它会读取我们的步骤,先给你看命令,等你同意。无需密钥,无需注册。 · 查看 AI 会读取的全文
通过 3/3 项检查 · 在你的电脑上运行 · ✓ 命令已检查 2026年11月19日
第一步:你用哪个 AI?
选择你用的应用,步骤会显示在下方。
可连接 Claude,但需要先准备
需要先在电脑上安装 Node.js(只需一次) · 仅限电脑上的 Claude 应用(Windows 或 Mac),网页和手机不可用
把这段话复制到 Claude 的对话里。AI 会先检查电脑,告诉你要改哪个文件,并且每次都等你同意。
请在这台电脑上为 Claude 安装 MCP server Sigvara,按顺序:1) 用 node -v 检查是否有 Node.js。没有的话,请让我从 https://nodejs.org/en/download 安装,不要自己安装。2) 只使用这份配置(已通过 PopNx MCP 检查):{ "mcpServers": { "sigvara": { "command": "npx", "args": [ "-y", "@sigvara/protocol-sdk" ] } } } 3) 告诉我要修改的文件完整路径或要运行的命令,等我同意。4) 不需要密钥。5) 按约定执行,然后确认 Sigvara 的工具真的出现了。这段话不包含你的密钥。如果需要密钥,AI 会请你自己粘贴到文件或应用的输入框里。
已经装好了吗?先检查 · 打开终端,输入这条命令后按 Enter
node -vMac:按 ⌘ + Space,输入 Terminal 后按 Enter
Windows:打开“开始”菜单,输入 Windows PowerShell,然后选“打开”
还没安装: 打开官方下载页,选择 LTS 版本和点选式安装程序:Mac 用 .pkg · Windows 用 .msi。双击后一路下一步即可,不需要 curl 命令
装好后关掉终端再打开,重新运行同一条命令,应显示版本号: node -v
安装一次,所有需要 Node.js 的工具都能用
SettingsDeveloperEdit Config
{
"mcpServers": {
"sigvara": {
"command": "npx",
"args": [
"-y",
"@sigvara/protocol-sdk"
]
}
}
}用 Sigvara 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude 说正在使用 Sigvara,或回答中出现 Sigvara 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
ChatGPT 暂时无法连接 Sigvara
Sigvara 需要安装在你的电脑上,而 ChatGPT 只能连接在网上运行的服务器
A · 使用能连接 Sigvara 的应用
B · 其他 类中能连接 ChatGPT 的其他服务器
从同类中使用较多的里挑选,仅限用地址即可连接、无需密钥的
可连接 Claude Code,但需要先准备
需要先在电脑上安装 Node.js(只需一次)
把这段话复制到 Claude Code 的对话里。AI 会先检查电脑,告诉你要改哪个文件,并且每次都等你同意。
请在这台电脑上为 Claude Code 安装 MCP server Sigvara,按顺序:1) 用 node -v 检查是否有 Node.js。没有的话,请让我从 https://nodejs.org/en/download 安装,不要自己安装。2) 只使用这份配置(已通过 PopNx MCP 检查):claude mcp add --scope user sigvara -- npx -y @sigvara/protocol-sdk 3) 告诉我要修改的文件完整路径或要运行的命令,等我同意。4) 不需要密钥。5) 按约定执行,然后确认 Sigvara 的工具真的出现了。这段话不包含你的密钥。如果需要密钥,AI 会请你自己粘贴到文件或应用的输入框里。
已经装好了吗?先检查 · 打开终端,输入这条命令后按 Enter
node -vMac:按 ⌘ + Space,输入 Terminal 后按 Enter
Windows:打开“开始”菜单,输入 Windows PowerShell,然后选“打开”
还没安装: 打开官方下载页,选择 LTS 版本和点选式安装程序:Mac 用 .pkg · Windows 用 .msi。双击后一路下一步即可,不需要 curl 命令
装好后关掉终端再打开,重新运行同一条命令,应显示版本号: node -v
安装一次,所有需要 Node.js 的工具都能用
Mac:按 Command+空格,输入 Terminal · Windows:打开 PowerShell
claude mcp add --scope user sigvara -- npx -y @sigvara/protocol-sdk应能看到 Sigvara 在列表中并显示已连接
用 Sigvara 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude Code 说正在使用 Sigvara,或回答中出现 Sigvara 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
可连接 Cursor,但需要先准备
需要先在电脑上安装 Node.js(只需一次)
把这段话复制到 Cursor 的对话里。AI 会先检查电脑,告诉你要改哪个文件,并且每次都等你同意。
请在这台电脑上为 Cursor 安装 MCP server Sigvara,按顺序:1) 用 node -v 检查是否有 Node.js。没有的话,请让我从 https://nodejs.org/en/download 安装,不要自己安装。2) 只使用这份配置(已通过 PopNx MCP 检查):{ "mcpServers": { "sigvara": { "command": "npx", "args": [ "-y", "@sigvara/protocol-sdk" ] } } } 3) 告诉我要修改的文件完整路径或要运行的命令,等我同意。4) 不需要密钥。5) 按约定执行,然后确认 Sigvara 的工具真的出现了。这段话不包含你的密钥。如果需要密钥,AI 会请你自己粘贴到文件或应用的输入框里。
已经装好了吗?先检查 · 打开终端,输入这条命令后按 Enter
node -vMac:按 ⌘ + Space,输入 Terminal 后按 Enter
Windows:打开“开始”菜单,输入 Windows PowerShell,然后选“打开”
还没安装: 打开官方下载页,选择 LTS 版本和点选式安装程序:Mac 用 .pkg · Windows 用 .msi。双击后一路下一步即可,不需要 curl 命令
装好后关掉终端再打开,重新运行同一条命令,应显示版本号: node -v
安装一次,所有需要 Node.js 的工具都能用
用 Sigvara 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Cursor 说正在使用 Sigvara,或回答中出现 Sigvara 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Cursor 帮助页面核实 · 你的界面可能因版本略有不同
可连接 VS Code,但需要先准备
需要先在电脑上安装 Node.js(只需一次)
把这段话复制到 VS Code 的对话里。AI 会先检查电脑,告诉你要改哪个文件,并且每次都等你同意。
请在这台电脑上为 VS Code 安装 MCP server Sigvara,按顺序:1) 用 node -v 检查是否有 Node.js。没有的话,请让我从 https://nodejs.org/en/download 安装,不要自己安装。2) 只使用这份配置(已通过 PopNx MCP 检查):{ "servers": { "sigvara": { "type": "stdio", "command": "npx", "args": [ "-y", "@sigvara/protocol-sdk" ] } } } 3) 告诉我要修改的文件完整路径或要运行的命令,等我同意。4) 不需要密钥。5) 按约定执行,然后确认 Sigvara 的工具真的出现了。这段话不包含你的密钥。如果需要密钥,AI 会请你自己粘贴到文件或应用的输入框里。
已经装好了吗?先检查 · 打开终端,输入这条命令后按 Enter
node -vMac:按 ⌘ + Space,输入 Terminal 后按 Enter
Windows:打开“开始”菜单,输入 Windows PowerShell,然后选“打开”
还没安装: 打开官方下载页,选择 LTS 版本和点选式安装程序:Mac 用 .pkg · Windows 用 .msi。双击后一路下一步即可,不需要 curl 命令
装好后关掉终端再打开,重新运行同一条命令,应显示版本号: node -v
安装一次,所有需要 Node.js 的工具都能用
用 Sigvara 告诉我你能帮我做什么,然后做一件给我看✓ 如果 VS Code 说正在使用 Sigvara,或回答中出现 Sigvara 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 VS Code 帮助页面核实 · 你的界面可能因版本略有不同
Claude Code
claude mcp add --scope user sigvara -- npx -y @sigvara/protocol-sdkClaude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"sigvara": {
"command": "npx",
"args": [
"-y",
"@sigvara/protocol-sdk"
]
}
}
}Cursor (~/.cursor/mcp.json)
{
"mcpServers": {
"sigvara": {
"command": "npx",
"args": [
"-y",
"@sigvara/protocol-sdk"
]
}
}
}VS Code (mcp.json)
{
"servers": {
"sigvara": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@sigvara/protocol-sdk"
]
}
}
}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。
依据这个服务器的连接方式和各应用自己的文档,检查于 2026年11月10日。
在你的电脑上运行,适用于
ClaudeGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatRaycast
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19日
注册名称已通过 GitHub 账户 RunTimeAdmin 验证,该账户拥有此仓库
代码最后更新于昨天
使用 MIT
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
尚未检查
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/servers/runtimeadmin-sigvara)提问、分享技巧或报告无法使用。按“有帮助”数量排序。
还没有评价
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 3 项适用的基础检查中通过了 3 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 仓库最后更新于昨天,最新版本(1.0.0-alpha.22)于今天进入注册中心。
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:World Monitor, openstatus, Oh My Posh Validator, Strata, Jitsu
同一分类、其他作者、命令已通过检查。官方的排在前面,再按星标排序。
如何看卡片:↓ 每周下载量i★ 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
Real-time global intelligence on conflict, markets, energy, trade, shipping and cyber. 100 tools.
Manage monitors, status pages, incidents and maintenances in your openstatus workspace.
Validate oh-my-posh configurations and segment snippets against the official schema.
MCP server for progressive tool usage at any scale (see https://klavis.ai)
Manage Jitsu data pipelines: destinations, streams, connections, functions, live events.
| 本项 Sigvara | World Monitor | openstatus | Oh My Posh Validator | Strata | Jitsu | |
|---|---|---|---|---|---|---|
| 运行位置 | 需要安装在电脑上 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 |
| 需要密钥 | 否 | 否 | 否 | 否 | 否 | 否 |
| 可在 Claude 中按 + 添加 | 否 | 否 | 否 | 否 | 否 | 否 |
| 基础检查 | 3/3 | 4/4 | 4/4 | 4/4 | 1/4 | 4/4 |
| 许可证 | MIT | AGPL-3.0 | AGPL-3.0 | MIT | Apache-2.0 | MIT |
| 价格 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 |
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。
TypeScript SDK for the Sigvara protocol: on-chain agent identity, computed reputation, staked accountability, and Ed25519 challenge-response authentication between agents. Successor to @countersig/protocol-sdk.
npm install @sigvara/protocol-sdk
Runtime dependencies are ethers v6 and tweetnacl. Node 18 or newer.
Use 1.0.0-alpha.12 or later. On 1.0.0-alpha.9 and earlier an agent object carries its secret key in an ordinary property, so JSON.stringify(agent) or logging one includes it; never log or serialize an agent from those versions. If one was logged, serialized or sent to an error reporter, treat its key as exposed and rotate it (Key Rotation).
Upgrading from 1.0.0-alpha.11:
agent.signChallengeis removed. It signed whatever it was
handed, so a verifier could get an agent to sign a challenge naming a different verifier and
present the signature there as the agent. A call does not compile in TypeScript, and in JavaScript
it throws an error that names
signChallengeFor(payload, audience), which signs only a challenge
addressed to the audience you pass and works from 1.0.0-alpha.10. The exported function
signChallenge(payload, secretKey)is unchanged. Details in CHANGELOG.md.
Upgrading from 1.0.0-alpha.10:
verifySignaturenow requires the audience, as its fifth
argument. Leaving it out used to compile and skip the check of who the agent had signed for, so a
response given to one verifier was accepted at another, and a v1 payload was accepted too. It now
does not compile in TypeScript, throws
expectedAudience is requiredin JavaScript, and a v1 payload
is always refused. A caller that already passes the audience needs no change. Details in
Upgrading from 1.0.0-alpha.9: the secret key is now a private field, so code that read it by
reaching around the type system stops working; sign with
signChallengeFor. It signs only a
challenge addressed to the audience you pass, where
signChallengesigned whatever it was handed
and is removed in 1.0.0-alpha.12. A gate now also checks the bond held now, so an agent
admitted at an old minimum stake and not topped up is refused as
not_active. It refuses to be
built with a
thresholdorttlSecondsthat is not a usable number, refuses a challenge dated
more than 60 seconds ahead, and gives up on a chain read after 30 seconds (
timeoutMs). The
optional
maxScoreAgeSecondsrefuses a score nobody has refreshed. Details in
Upgrading from 1.0.0-alpha.7:
generateChallengenow takes an audience as its
second argument, and every existing call site needs it.
`diff
- const challenge = generateChallenge(agentDid);
+ const challenge = generateChallenge(agentDid, 'https://your-service.example');
`
Without it, a verifier holding a valid response could present that same response to a
different verifier and be accepted as the agent. Pass the same audience to
verifySignatureas its fifth argument; on 1.0.0-alpha.10 and earlier, omitting it still
compiles and still accepts the relayable older proofs. Full detail in CHANGELOG.md.
This does not make you replay-proof. The audience stops a response being reused at
a *different* verifier. Reuse at *yours* is stopped by remembering the nonce, and the
SDK cannot do that for you — see
What the SDK still needs from you.
did:sigvara: : and its on-chain didHash, and register the agent from the operator wallet, with a proof of control signed by the agent address.meetsThreshold, stake and minimum-stake checks, and a W3C DID Document built from chain state.depositStake handles the ERC-20 approval and the deposit in one call, and is what moves a newly registered agent from PendingBond to Active.SigvaraGate turns the reads above into one decision: issue a challenge, admit or refuse, and say which reason applies. It owns the nonce, so replay protection is not left to the caller.SigvaraTrustCheck asks the gate's question about a bare address. sigvaraX402Fetch puts it in front of an x402 payment, and sigvaraTrustTool and requireTrustedAgent put it in an agent framework's tools.signPaymentOutcome), so the oracle can count it without a token, and check such a signature the way the oracle does (verifyPaymentOutcome).Scores come in two forms. The earned score is what the oracle last finalized; the total score is the matured one, which climbs toward it over time and is what meetsThreshold checks. A newly bonded agent has earned a score before it can spend it.
import { SigvaraAgent, SigvaraVerifier, registerAgent, depositStake, generateChallenge } from '@sigvara/protocol-sdk';
import { ethers } from 'ethers';
const RPC = 'https://rpc.testnet.arc.io'; const CHAIN_ID = 5042002; // Arc testnet const addresses = { identity: '0x…', reputation: '0x…', staking: '0x…' }; // deployments/5042002.json
// Operator side: create, register and bond an agent const { agent, privateKey } = SigvaraAgent.generate({ agentAddress: '0xAgent…', chainId: CHAIN_ID }); const operator = new ethers.Wallet(process.env.OPERATOR_KEY!, new ethers.JsonRpcProvider(RPC));
// The agent address must prove control of itself. Pass a signer for it, or // { signature } if it was signed elsewhere (HSM, Safe, any ERC-1271 contract). const agentSigner = new ethers.Wallet(process.env.AGENT_KEY!); // a signer for agent.agentAddress await registerAgent(operator, agent.agentAddress, agent.publicKeyBytes32, addresses.identity, { agentSigner });
// Registration leaves the agent PendingBond. This deposit is what activates it. await depositStake(operator, agent.didHash, 1000n * 10n ** 18n, addresses.staking);
// Counterparty side: challenge the agent, verify the signature, read status and score const verifier = new SigvaraVerifier({ rpcUrl: RPC, addresses, chainId: CHAIN_ID }); // The second argument is the audience: who is asking. Pass a stable identifier you // control, normally your own origin. It is signed, so the agent's response proves it was // talking to *you* and cannot be relayed onward by anyone who receives it. const challenge = generateChallenge(agent.did, 'https://your-service.example'); // Done by the agent, returned to the counterparty. signChallengeFor refuses a challenge that is not // addressed to the verifier the agent means to prove itself to, which is what stops a verifier from // collecting a signature meant for someone else. const signature = agent.signChallengeFor(challenge.payload, 'https://your-service.example');
// Pass the same audience back when verifying. It is required: verifySignature throws without // it, and it refuses a v1 payload, which names no audience to check. const ok = await verifier.verifySignature( agent.did, challenge.payload, signature, 300, 'https://your-service.example', ); const trusted = ok && (await verifier.isActive(agent.did)) && (await verifier.meetsThreshold(agent.did, 40));
The 40 in these examples is a placeholder. A score below roughly 60 says little on its own, so read what a fake score costs before you pick a number, and sigvara.xyz/status for how the scores spread today.
Signature validity, status and score are separate reads. Combining them correctly is easy to get subtly wrong, so there is a piece that does it.
SigvaraGateimport { SigvaraGate } from '@sigvara/protocol-sdk';
const gate = new SigvaraGate({ rpcUrl: RPC, addresses, chainId: CHAIN_ID, threshold: 40, audience: 'https://your-service.example', // required, and signed into every challenge });
// 1. hand the challenge to the agent const challenge = gate.challenge(agentDid);
// 2. the agent signs challenge.payload and returns the signature
// 3. decide const result = await gate.admit(agentDid, challenge, signature); if (!result.ok) { // 'bad_proof' | 'replayed' | 'not_active' | 'below_bond' | 'stale_score' | 'below_threshold' return refuse(result.reason); }
The reasons are separate because they need different responses. bad_proof means nothing was proven. replayed means a valid signature arrived twice. not_active means a real, proven agent that is unbonded, suspended or slashed, and telling that caller to earn more points would be wrong. "Unbonded" is the bond held now against the minimum stake in force now, not the status stored when the agent was admitted: a minimum raised since leaves an agent marked Active while it holds less than the contract asks, and the gate reads the bond for that reason. below_bond appears only if you set minBond, and stale_score only if you set maxScoreAgeSeconds. Only below_threshold is about reputation, and it carries the score so you can say how short they were.
A score nobody has refreshed. Nothing on chain ages a score, so one left over from an oracle outage keeps passing a threshold. Set maxScoreAgeSeconds to refuse a score written longer ago than that, or never, with stale_score. A healthy score gets to about two of your oracle's proposals apart before the next one lands (about 20 hours on Arc testnet, where an unchanged score is proposed again every 12 hours, two days on a daily epoch), so choose more than that. It is measured against your host's clock. Unset, a score is trusted however old it is. It costs one more read per admission. The same check exists on chain as SigvaraReader.isTrustedFresh, on Arc testnet at 0x2921b4eE879CdF5793FA273BbDc4E7dF01ABcef8.
More at stake than the minimum covers. Set minBond to refuse an agent holding less, with below_bond, whatever its score (from 1.0.0-alpha.16). It is in the bond token's base units, as a bigint or a string of digits. A number is refused, because a floor past what a double can hold would quietly become another floor. A floor under the protocol's minimum still needs the minimum, and an amount queued for withdrawal does not count. It compares the same number as SigvaraReader.isTrusted(agent, minScore, minBond), and costs one more read per admission.
The constructor refuses a threshold that is not a finite number of 0 or more, and a ttlSeconds or maxScoreAgeSeconds that is not a positive number. A NaN threshold, for example Number(process.env.THRESHOLD) with the variable unset, would otherwise admit every active agent that can prove its key, whatever its score.
A gate that has to say how it runs. Two of the defaults suit a test bench and not a deployment, and both fail open by being left out: instances defaults to one process, and no maxScoreAgeSeconds means no age limit. profile: 'production' (from 1.0.0-alpha.14) makes the constructor refuse to build the gate until both are set, and names everything missing in one error:
const gate = new SigvaraGate({
rpcUrl: RPC, addresses, chainId: CHAIN_ID, threshold: 40, audience: 'https://your-service.example', profile: 'production', instances: 'multiple', // or 'single' if exactly one process serves this gate nonceStore, // shared and atomic, required with 'multiple' (see below) maxScoreAgeSeconds: 86_400, // more than two of your oracle's