原型 · 预览版:收录官方注册中心里的全部服务器,每 30 分钟同步一次。使用 Google 登录即可撰写评价。原型 · 每 30 分钟更新
PopNx MCP为使用 AI 的人准备的 MCP 服务器 · 用大白话讲清楚

Graviti

✓ thesingulariti.ai

作者:sclaytonmyosin版本 0.2.0i★ 0i在云端运行官方网站 ↗
注册名称ai.thesingulariti/graviti

安装方法

What it does

Verified brand claims with receipts for agent commerce. Ranking is never paid; the engine is open.

  • Connects 在云端运行
  • Account or key No key needed
  • Category 网店与支付

为什么排在这里

此项在“连接 AI 的应用”标签中,该标签不使用星标,按以下顺序排列

1. 在 Claude 连接器目录中否
2. 团队实测过否
3. 适合泰国用户否
4. 不计星标的使用量15/100热度 0 · 新鲜度 7 · 仍在响应 8 · 用户喜欢 +0
5. 登记日期2026-08-26

用户喜欢: 0 · 最多加 5 分(每人 1 分),不能用钱买。

质量: 基础检查通过 4/5 项(仅展示,不用于排序)

安装

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 Graviti。按照 https://mcp.popnx.com/i/thesingulariti-graviti.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 4/5 项检查 · 你的提问会经过开发者的服务器(graviti.thesingulariti.ai) · ✓ 命令已检查 2026年11月13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 · 约 2 分钟

不需要密钥,不用安装程序 · 在 claude.ai 或桌面应用设置一次,即可在聊天中使用 · 免费账户可添加 1 个 · Team/Enterprise 方案由所有者(Owner)添加

  1. 打开 Claude 的 Connectors 页面

    CustomizeConnectors

    CustomizeConnectors1Add2https://…3Add4
  2. 点击“Add” → “Custom” → “Web”,填写这 2 项

    把这里的内容复制到 Claude 中同名的输入框,然后点击 Continue

    Name · 名称Graviti
    Remote MCP server URL · 地址https://graviti.thesingulariti.ai/api/mcp

    在 Claude 检测到的登录设置页点击 Continue · Authentication 按作者说明选择(Sign in now、Sign in when needed 或 No sign in),然后点击 Add

  3. 在聊天中开启

    +ConnectorsGraviti

  4. 试一试

    用 Graviti 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 Graviti,或回答中出现 Graviti 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

能否连接 ChatGPT 取决于方案

完整 MCP:网页版 ChatGPT Business、Enterprise 或 Edu · Pro:以读取/获取权限连接,修改数据的工具可能用不了 · Plus:官方文档尚未确认,如果在插件页看到添加自定义 MCP 的选项可以试试 · 仅限网页

  1. 在网页版 ChatGPT 打开添加 MCP 的页面

    Plugins+Add custom MCP server

    ChatGPTPlugins1+2https://…3Create4
  2. 填写名称和地址

    输入名称,在 Connection 下粘贴地址,勾选风险提示,然后点击 Create as a plugin

    Name · 名称Graviti
    Connection · 地址https://graviti.thesingulariti.ai/api/mcp
  3. 在聊天中选择 Graviti

    开启新聊天,输入前先从插件列表中选择 Graviti

  4. 试一试

    用 Graviti 告诉我你能帮我做什么,然后做一件给我看

    ✓ 如果 ChatGPT 说正在使用 Graviti,或回答中出现 Graviti 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 OpenAI 帮助页面核实 · 你的界面可能因版本略有不同

可连接 Claude Code · 约 2 分钟

不需要密钥,不用安装程序

  1. 打开终端

    Mac:按 Command+空格,输入 Terminal · Windows:打开 PowerShell

  2. 粘贴这条命令并按 Enter

    claude mcp add --transport http --scope user graviti https://graviti.thesingulariti.ai/api/mcp
  3. 在 Claude Code 中输入 /mcp

    应能看到 Graviti 在列表中并显示已连接

  4. 试一试

    用 Graviti 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 Graviti,或回答中出现 Graviti 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 · 约 2 分钟

不需要密钥,不用安装程序

  1. 点击这个按钮打开 Cursor

    在 Cursor 中打开

    Cursor 会请你确认安装,点击 Install

  2. 在 Cursor 中以 Agent 模式打开 AI 聊天

  3. 试一试

    用 Graviti 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 Graviti,或回答中出现 Graviti 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 · 约 2 分钟

不需要密钥,不用安装程序

  1. 点击这个按钮打开 VS Code

    在 VS Code 中打开

    VS Code 会请你确认安装,点击 Install

  2. 在 VS Code 中以 Agent 模式打开 AI 聊天

  3. 试一试

    用 Graviti 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 Graviti,或回答中出现 Graviti 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

其他支持 MCP 的应用(如 Windsurf、Zed、Gemini CLI)添加方式各不相同,请在应用列表中查看

查看所有支持 MCP 的应用 →

通过网页连接的应用使用的地址

地址https://graviti.thesingulariti.ai/api/mcp

JSON 设置在下方“开发者专用”中

卡住了?问问正在用 Graviti 的人💬 讨论 →
开发者专用:终端命令和 JSON 设置

通过网页连接

Claude Code

claude mcp add --transport http --scope user graviti https://graviti.thesingulariti.ai/api/mcp

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "graviti": {
      "url": "https://graviti.thesingulariti.ai/api/mcp"
    }
  }
}

VS Code (mcp.json)

{
  "servers": {
    "graviti": {
      "type": "http",
      "url": "https://graviti.thesingulariti.ai/api/mcp"
    }
  }
}

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。

一览

每周下载量i
不在 npm/PyPI 上
已验证域名
✓ thesingulariti.ai
许可证i
Apache-2.0
连接方式
在云端运行
传输方式
streamable-http
最新版本i
0.2.0
发布到注册中心i
2026年8月26日
代码最后更新i
2026年9月7日
最近检查
2026年11月15日
作者
sclaytonmyosin
注册状态
活跃

哪些应用可以使用它

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。

通过互联网连接,适用于

ClaudeChatGPTGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatOpen WebUIRaycast

全部应用及其优缺点 →

它能做什么

工具名称和描述与服务器在 2026年11月15日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。

  • match_intent

    Takes a natural-language buyer question plus optional intent profile; returns recommendations ranked purely by fit score. Verification status is disclosed but never affects rank. Questions outside the verified index return an explicit out-of-scope result (in_index: false), never

  • get_verified_claims

    Structured claims for one brand: provenance URLs, freshness timestamps, corroboration counts, and verification status (verified means the brand pays for auditing — always disclosed). Where captured, includes the specifications block (ingredient/spec disclosure with per-fact prove

  • category_landscape

    Comparison-table view of every brand in a category across disclosed attributes, with verification status marked per entry. Entries with captured ingredient/spec data also carry info_quality (0–100 disclosure completeness — transparency, never a rank factor; absent means not yet a

  • get_accountability_log

    Public, machine-readable log of every seal lifecycle change: flags, degradations, revocations, and restorations, with dates and reasons — plus confirmation-loop events (event: confirmation_request / confirmation_resolved), where the audit asked a brand to confirm a claim it could

  • get_ledger

    Append-only version history of the index: each entry commits to the SHA-256 of the canonical index content and the previous entry's hash (rewriting history breaks every hash after it), is Ed25519-signed, and is anchored to Bitcoin via OpenTimestamps. Returns entries plus independ

  • get_gap_report

    What stands between a brand and top-of-category: attribute gaps vs the category leader, open flags, cohort blind spots per buyer-intent voice, and the ingredient/spec disclosure dimension — the brand's info_quality score with the exact publishable fixes that would raise it (publi

  • report_conversion

    Appends one attributed-transaction event to a log — the demo stub of the pay-on-real-acquisition loop. It moves no money, calls no external service, and never deletes or overwrites anything; the response states where (and whether) the event was persisted on this transport. Paymen

基础检查i

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

  • ! 注意

    作者与仓库一致

    发布者(thesingulariti.ai)不是仓库所有者(sclaytonmyosin);可能是其他人代为上架的

  • ✓ 通过

    仍在维护

    代码最后更新于2个月前

  • ✓ 通过

    有许可证

    使用 Apache-2.0

  • ✓ 通过

    端点有响应

    在 540 毫秒内响应(HTTP 200)

  • – 不适用

    使用 OAuth 登录

    无需登录

  • ✓ 通过

    工具描述中没有隐藏指令

    已读取 7 个工具描述,未发现针对 AI 的隐藏指令

我们检查什么 → · JSON

作者:申请重新检查或报告信息错误
给作者:在 README 中加徽章

如果你是作者,可把这段代码贴到 README。徽章会链接回本页。

[![Featured on PopNx MCP](https://mcp.popnx.com/badge/thesingulariti-graviti.svg)](https://mcp.popnx.com/en/servers/thesingulariti-graviti)

讨论

提问、分享技巧或报告无法使用。按“有帮助”数量排序。

还没有帖子,欢迎第一个提问或分享。

此处内容由用户撰写,不代表 PopNx 团队。

登录后发帖

用户评价

还没有评价

登录后写评价

与“网店与支付”中的其他服务器比较i

  • 星标数多于此分类中其他 2941 个服务器的 0%
  • 发布到注册中心的时间比此分类中其他 2941 个服务器的 23% 更新
  • 基础检查结果等于或优于此分类中其他 2941 个服务器的 63%

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。

常见问题

Graviti 安全吗?

它在 5 项适用的基础检查中通过了 4 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。

它需要哪些密钥或代码?

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。

它在我的电脑上运行还是在云端?

它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。

它还在维护吗?

GitHub 仓库最后更新于2个月前,最新版本(0.2.0)于3个月前进入注册中心。

哪些应用可以使用它?

支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,因此也能用于接受远程 MCP 的应用,例如网页版 ChatGPT(Business 及以上为完整 MCP;Pro 为只读)。

有什么可以替代它?

同一分类中有:Stripe MCP, flightclaw, Trade It, Marketplace Ad Pros: Amazon Ads & Seller Central, idealo MCP Server

功能相近的其他选择

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

Stripe MCP

MCP server integrating with Stripe - tools for customers, products, payments, and more.

作者:stripe广泛使用★ 207213个月前更新

在云端运行✓ 服务方 · stripe.com

flightclaw

Search ~300 airlines and book flights with a checkout link. OAuth sign-in by email code.

作者:flightclaw正在增长★ 862个月前更新

在云端运行✓ 服务方 · flightclaw.com

Trade It

Connect brokerage accounts to AI assistants: holdings, orders, asset lookup, and approved trading.

作者:trade-it-inc起步阶段★ 6234天前更新

在云端运行✓ 服务方 · tradeit.app

idealo MCP Server

Search products, compare prices and discover deals across 6 European markets with your AI assistant.

作者:idealo起步阶段★ 286个月前更新

在云端运行✓ 服务方 · mcp.idealo.com
与相似项目对比
本项
Graviti
Stripe MCPflightclawTrade ItMarketplace Ad Pros: Amazon Ads & Seller Centralidealo MCP Server
运行位置在云端运行,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 + 添加否是否否否否
基础检查4/54/44/45/54/44/4
许可证Apache-2.0MITMITCC0-1.0MITMIT
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

作者的 README

打开 README

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。

graviti-engine

The deterministic scoring engine behind Graviti — trust infrastructure for agent-mediated commerce, built by Singulariti. This repository exists so the site's central claim ("scoring is open to audit") is checkable, not just stated: src/engine.ts here is the exact module that ranks every recommendation in production — the MCP servers, the web console, and the public endpoints all import this one file. There is no private ranking path.

Use it from any agent — under a minute

Graviti speaks MCP on two paths; both serve the same seven tools (match_intent, get_verified_claims, category_landscape, get_accountability_log, get_ledger, get_gap_report, report_conversion) through this exact engine.

Remote (fastest) — hosted Streamable HTTP endpoint, no auth, no account:

{ "mcpServers": { "graviti": { "url": "https://graviti.thesingulariti.ai/api/mcp" } } }

Local — the graviti-mcp npm package over stdio:

{ "mcpServers": { "graviti": { "command": "npx", "args": ["-y", "graviti-mcp"] } } }

The package ships an index snapshot and, on startup, pulls the latest published index and verifies it against the signed ledger — the pinned key, two independent ledger paths, and the replay guard are the same procedure as scripts/verify-payload.mjs below. A tampered or replayed payload is never used. (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; wiring for other clients at graviti.thesingulariti.ai/use-graviti.)

The two laws

  1. Verification is paid. Ranking is never paid. Brands pay for continuous independent auditing of their claims. Verification status carries zero weight in ranking — read the fit-score construction in matchIntent() and confirm no payment signal enters it.
  2. Payment buys speed, never outcome. A brand can pay to fast-track a re-evaluation; payment compresses the clock, never the verdict.

The only score adjustments tied to a brand's standing are the fixed, disclosed lifecycle penalties (LIFECYCLE_PENALTY: flagged −0.05, degraded −0.15) — driven by monitored quality signals, never by payment status — and revoked brands are excluded from scoring entirely.

What you can verify by reading this code

  • Scoring is deterministic. Pure functions over a brand index — no I/O, no network, no randomness, no learned parameters. The engine's only inputs are the index, the question text, and an optional intent profile. Same inputs, same scores, every time.
  • Conversion data never enters scoring. The engine has no conversion input — grep this file for anything conversion-shaped and you'll find nothing. Attributed-conversion events exist in the private transport layer purely as billing records (pay-on-real-acquisition); they are structurally incapable of reaching a fit score because this module never sees them.
  • Scores are reproducible from public data. The full index this engine runs over in production is public. Point the engine at it and you reproduce production scores exactly:
  • https://graviti.thesingulariti.ai/index.json — manifest (totals, disclosure, category list)
  • https://graviti.thesingulariti.ai/index-full.json — the complete index in one fetch; this is the exact content the integrity ledger hashes
  • https://graviti.thesingulariti.ai/ledger.json — append-only, hash-chained, Ed25519-signed, Bitcoin-anchored version history
  • Every payload carries structural disclosure. See disclosure(): rank_influenced_by_payment: false, who paid, who didn't, and what payment buys — machine-readable, on every response.
  • Out-of-scope questions get a refusal, never a guess. See the in_index: false branch of matchIntent().

Run it

npm install

npm run demo -- "best magnesium supplement" npm run demo -- "cheapest cold plunge with a chiller" npm run demo -- "best magnesium supplement" --json # the full machine payload agents receive

The demo runs the engine over data/sample-index.json, a small real slice of the public index (five categories, ~100 brands, with claims, provenance URLs, and the accountability log). For the complete picture, fetch index-full.json and substitute it — the engine takes any BrandIndex.

Run the canaries

npm run canary

Graviti runs canary-based drift detection: subjects of known ground truth are pushed through the system on every production build (Panel A — golden routing queries, structural invariants like *no NaN score anywhere* and *catalog-tier brands never rank*, and adversarial cases where the famous answer must NOT win) and through the audit crawler on every cycle (Panel B — three provenance pages the site controls: one steady, one deliberately drifted, one deliberately gone). A Panel A divergence fails the build; a Panel B misread means the sensor itself is broken and that crawl cycle refreshes no real claims. The two panels are independently maintained so neither becomes the other's ground truth.

scripts/canary-panel-a.ts is the public subset of Panel A — the cases the sample index supports, over the exact engine this repository publishes, so you can run the same canaries the build runs with zero secrets. Live status for both panels is published in the canaries block of pulse.json (human-readable at /pulse).

Verify the ledger

The index's version history is a hash chain: each entry commits to the SHA-256 of the canonical full-index content and the previous entry's hash, is Ed25519-signed against the published public key (data/public-key.pem, also served at /public-key.pem), and is anchored to Bitcoin via OpenTimestamps. Rewriting any historical entry breaks every hash after it — even Graviti cannot edit the record.

npm run verify:ledger           # verifies the committed ledger snapshot, then checks it against the live index

npm run verify:ledger -- --live # fetches and verifies the current live ledger instead

The script needs no secrets and no Graviti-run infrastructure: it recomputes every entry hash, walks the chain back to genesis, and checks every signature against the public key. Bitcoin anchors can be independently replayed with the open-source OpenTimestamps client against proofs at /ots/v .ots — instructions in llms.txt.

Verify what you received — transport integrity

The ledger proves what was *published*. It cannot see the pipe: TLS-terminating proxies, CDNs, caches, and load balancers sit between Graviti's origin and your process, and any of them could alter a payload with no ledger entry capturing the delta. From ledger v25, every entry's signed core pins a sha256 manifest of every published payload (files.artifacts), so the check moves to your side of the pipe. Verify before trust:

npm run verify:payload                                             # /index-full.json

npm run verify:payload -- /index.json /data/categories/magnesium-supplements.json /llms.txt

scripts/verify-payload.mjs is dependency-free (Node ≥ 20) and exits nonzero on any failure. It:

  1. fetches the ledger from two independent paths — this repo's snapshot on raw.githubusercontent.com and the origin's /ledger.json — verifies both chains, and cross-checks the heads (a fork or an origin serving an older ledger than the public mirror fails loudly);
  2. verifies every Ed25519 signature against the pinned signing key embedded in the script — the served /public-key.pem is cross-checked against the pin and a mismatch is a failure, never a source of trust;
  3. recomputes the sha256 of the payload you actually received and compares it to the head entry's signed manifest;
  4. guards against replay: a payload that matches an older entry's manifest instead of the head is stale-but-validly-signed — reported as its own failure, because a replayed yesterday is the subtle attack.

No Node? The core check is two commands with curl, jq, and sha256sum — expected hash from the mirror snapshot (independent path), actual hash from the bytes that reached you:

curl -s https://raw.githubusercontent.com/sclaytonmyosin/graviti-engine/main/data/ledger.json \

| jq -r '.entries[-1].files.artifacts["/index-full.json"]' curl -s https://graviti.thesingulariti.ai/index-full.json \ | jq -cjS 'del(.generated_at, .integrity)' | sha256sum

The two hashes must match. (JSON payloads are hashed in canonical form — recursively sorted keys, no whitespace — after deleting the two volatile envelope fields generated_at and integrity, which carry the build timestamp and the pinning entry's own hash and so cannot be inside the pinned content. /llms.txt is hashed as exact bytes, no jq step.) For the full guarantee, also confirm the mirror and origin ledgers agree and the signature chain verifies — that's steps 1–2 above, or npm run verify:ledger.

The boundary, honestly: a passing check proves the bytes that reached the edge of your process match Graviti's signed, Bitcoin-anchored record. Nothing cryptographic reaches *inside* your runtime or a model's context window after verification passes — if your own stack mutates the data afterwards, no publisher-side mechanism can see it. Verify as close to the point of use as you can; the last hop is yours.

Verify Graviti traffic — outbound attribution (for brands)

Every recommendation link Graviti produces to a brand's own site carries deterministic UTMs (utm_source=graviti, utm_medium= , utm_campaign= , utm_content= ) plus a signed gvt token, so the brand on the receiving end can *prove* the referral came from Graviti instead of trusting a spoofable UTM string. The token (gvt1. . . . . ) is Ed25519-signed with the same key that signs the ledger, and identifies the recommendation surface, never the user — no user identifiers, no click IDs, no per-click values. Full spec: docs/ATTRIBUTION.md.

node scripts/verify-attribution.mjs 'gvt1.agent.magnesium-supplements.bioptimizers.20260829.<sig>'

or paste the full landing URL — the script extracts the gvt parameter

Dependency-free (Node ≥ 20), pinned key, exit 0 only on a valid signature. The equivalent openssl commands are in the spec.

Sync provenance

src/engine.ts, scripts/verify-ledger.mjs (modulo the live-fetch adaptation noted in its header), data/ledger.json, and data/public-key.pem mirror the private Graviti monorepo at commit cf24943 (2026-08-26). The engine file is verbatim — re-syncing is a file copy plus an update to this line, and it now happens automatically with every ledger snapshot sync. This sync carries the rubric-pinning release: from ledger v26 every entry's signed core carries a rubric field — the sha256 and git blob sha1 of the exact committed src/engine.ts that scored every ranking published under that entry, plus the pinning private-repo commit and this repository's URL. That makes the sync provenance *checkable instead of stated*:

# The engine in this repo must hash to exactly what the signed ledger head pins:

git hash-object src/engine.ts curl -s https://graviti.thesingulariti.ai/ledger.json | jq -r '.entries[-1].rubric.engine_git_blob_sha1'

— the two must match byte for byte. Equivalently, with no git:

sha256sum src/engine.ts # must equal .entries[-1].rubric.engine_sha256

The blob hash is computed by git hash-object over the raw committed file bytes (no normalizat

只显示了开头部分;其余内容请在仓库中阅读