Vaultfire MCP Server
⚠️ Alpha 软件 — Vaultfire 协议正处于活跃开发阶段。智能合约已部署在主网上,但尚未经过第三方安全公司的正式审计。只读工具对任何代理都是安全的。写入工具会与实时合约交互,且交易不可逆。请自行承担风险。有关免责声明,请参阅 LICENSE。
@vaultfire/mcp-server
Vaultfire 协议 的 MCP 服务器
为任何兼容 MCP 的 AI 提供链上信任验证。 适用于 Claude · Copilot · Cursor · Windsurf · Google ADK 等。
部署在 Base · Avalanche · Arbitrum · Polygon 上。
快速入门
Claude Desktop
添加到你的 claude_desktop_config.json(在 macOS 上通常位于 ~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"vaultfire": {
"command": "npx",
"args": ["-y", "@vaultfire/mcp-server"],
"env": {
"VAULTFIRE_CHAIN": "base"
}
}
}
}要启用写入工具(注册代理、创建保证金),请添加你的私钥:
{
"mcpServers": {
"vaultfire": {
"command": "npx",
"args": ["-y", "@vaultfire/mcp-server"],
"env": {
"VAULTFIRE_CHAIN": "base",
"PRIVATE_KEY": "your_private_key_here"
}
}
}
}Cursor
添加到项目根目录下的 .cursor/mcp.json:
{
"mcpServers": {
"vaultfire": {
"command": "npx",
"args": ["-y", "@vaultfire/mcp-server"],
"env": {
"VAULTFIRE_CHAIN": "base"
}
}
}
}通用 (npx)
# Read-only mode (no key needed)
npx @vaultfire/mcp-server
# With a specific chain
VAULTFIRE_CHAIN=avalanche npx @vaultfire/mcp-server
# With write tools enabled
PRIVATE_KEY=your_private_key npx @vaultfire/mcp-server全局安装
npm install -g @vaultfire/mcp-server
vaultfire-mcp-server可用工具
只读工具(始终可用)
工具 | 描述 |
| 全面信任验证:身份、保证金、Street Cred、声誉、桥接状态 |
| 获取 Street Cred 分数 (0-95) 和等级 |
| 链上身份数据(URI、类型、注册日期、活跃状态) |
| 地址的所有合作伙伴保证金 |
| 声誉数据(平均评分、反馈数量、验证百分比) |
| 按能力标签查找代理 |
| 协议统计:总代理数、保证金、保证金价值、桥接同步 |
写入工具(需要 PRIVATE_KEY 环境变量)
工具 | 描述 |
| 在链上注册 AI 代理 (ERC-8004) |
| 创建带有经济抵押的合作伙伴保证金 |
警告: 写入工具会向主网提交真实交易。交易不可逆。
资源
服务器为协议上下文公开了两个 MCP 资源:
URI | 描述 |
| 所有 4 条支持的链,包含合约地址和链详情 |
| Street Cred 评分公式和等级定义 |
配置
环境变量 | 必需 | 默认值 | 描述 |
| 否 |
| 要查询的链: |
| 否 | — | 写入工具(注册、保证金)的私钥。切勿提交此项。 |
| 否 | — | 自定义 RPC URL(覆盖所选链的默认值) |
Street Cred 评分
Street Cred 是一个综合信任分数 (0–95),根据实时链上数据计算得出:
组件 | 分数 | 条件 |
身份已注册 | 30 | 代理拥有 ERC-8004 链上身份 |
有保证金 | 25 | 至少存在一个合作伙伴保证金 |
保证金活跃 | 15 | 至少有一个保证金当前处于活跃状态 |
保证金等级 | 0–20 | 白金 (≥0.5 ETH) / 黄金 (≥0.1) / 白银 (≥0.05) / 青铜 (≥0.01) |
多重保证金 | 5 | 拥有多个保证金(生态系统参与者) |
等级:
等级 | 分数范围 | 描述 |
未评级 | 0 | 无链上信任信号 |
青铜 | 1–30 | 基本身份或极少的保证金活动 |
白银 | 31–55 | 已注册并有一定抵押的保证金 |
黄金 | 56–75 | 活跃保证金且有可观抵押 |
白金 | 76–95 | 多个活跃保证金且有显著抵押 |
支持的链
链 | 链 ID | 中心 | 原生代币 |
Base | 8453 | ✅ 主要 | ETH |
Avalanche | 43114 | — | AVAX |
Arbitrum | 42161 | — | ETH |
Polygon | 137 | — | POL |
为什么选择 Vaultfire
特性 | Vaultfire | AxisTrust | Cred Protocol | Okta XAA |
AI 问责保证金 | ✅ | ❌ | ❌ | ❌ |
AI 合作伙伴保证金 | ✅ | ❌ | ❌ | ❌ |
链上、去信任化 | ✅ | ❌ | 部分 | ❌ |
多链(首日支持) | ✅ (4) | ❌ | ❌ | ❌ |
Street Cred 综合评分 | ✅ | T-Score | C-Score | ❌ |
信仰加权治理 | ✅ | ❌ | ❌ | ❌ |
符合 ERC-8004 | ✅ | ❌ | ✅ | ❌ |
道德高于指标。隐私高于监控。自由高于控制。 让促进人类繁荣比榨取价值更有利可图。
核心差异化优势:
AI 问责保证金 — 真正的经济利益绑定,而不仅仅是分数
AI 合作伙伴保证金 — 代理之间的相互经济利益绑定
信仰加权治理 — 行业首创(非代币加权)
首日即支持跨链 — 4 条主网链
Street Cred 评分 — 基于链上数据的 0-95 综合信任评分
Vaultfire 生态系统
包 | 描述 |
核心 SDK — 注册代理、创建保证金、查询声誉 | |
LangChain / LangGraph 集成 | |
代理对代理 (A2A) 协议桥接 | |
企业级 IAM 桥接 (Okta, Azure AD, OIDC) | |
本包 — 用于 Claude、Copilot、Cursor 的 MCP 服务器 | |
OpenAI Agents SDK 集成 | |
Vercel AI SDK 中间件和工具 | |
带有信任验证的 XMTP 消息传递 | |
带有信任门控的 X402 支付协议 | |
Vaultfire 命名服务 — 人类可读的代理 ID | |
CrewAI 集成 (Python) | |
3 个具有实时链上信任的参考代理 | |
A2A 信任扩展规范 — 代理卡的链上信任 | |
为什么 Vaultfire 保证金优于信任分数 — 实时证明 | |
信任框架白皮书 — AI 的经济问责制 | |
开发者门户 — 快速入门、演练场、框架选择器 |
安全性
只读工具无需凭据,在任何地方运行都是安全的
写入工具需要
PRIVATE_KEY— 请始终从环境变量加载,切勿硬编码智能合约已部署在主网上,但尚未经过正式审计
通过写入工具发送的交易不可逆
请勿向不受信任的方公开你的 MCP 服务器配置(包含
PRIVATE_KEY)
许可证
MIT — 版权所有 2025 Ghostkey316
有关完整条款和免责声明,请参阅 LICENSE。
Available Tools
7 toolsvaultfire_discover_agentsC
Find registered AI agents by capability tags. Returns a list of agent addresses that match the requested capabilities. Capabilities are hashed on-chain.
Inputs: capabilities (required), chain.
vaultfire_get_agentA
Get on-chain identity data for an agent address: agent URI, agent type, registration timestamp, and active status. Returns ERC-8004 compliant identity info.
Inputs: address (required), chain.
vaultfire_get_bondsB
Get all partnership bonds for an address. Returns bond IDs and full bond details including partner addresses, partnership type, stake amount (in ETH/native token), creation date, and active status.
Inputs: address (required), chain.
vaultfire_get_reputationB
Get reputation data for an agent: average rating, total feedback count, verified feedback count, and percentage of verified feedback. Ratings are stored on-chain from real interactions.
Inputs: address (required), chain.
vaultfire_get_street_credA
Get the Street Cred score (0-95) and tier for an agent address. Tiers: Unranked (0), Bronze (1-30), Silver (31-55), Gold (56-75), Platinum (76-95). Also returns the score breakdown showing how each component contributes.
Inputs: address (required), chain.
vaultfire_protocol_statsA
Get Vaultfire Protocol statistics: total registered agents, total bonds created, total active bond value (in native token), and bridge sync count.
Inputs: chain.
vaultfire_verify_agentA
Full trust verification for an AI agent address: checks on-chain identity (ERC-8004), partnership bonds, Street Cred score (0-95), reputation data, and cross-chain bridge recognition. Returns a trusted/untrusted verdict with reason. Use this before interacting with an unknown agent.
Inputs: address (required), chain, minScore.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 7 tools
Each tool has a clearly distinct purpose with no overlap: discover_agents finds agents by tags, get_agent retrieves identity data, get_bonds lists partnership bonds, get_reputation provides rating data, get_street_cred gives a score and tier, protocol_stats shows system-wide metrics, and verify_agent performs a comprehensive trust check. The descriptions reinforce these unique functions, making misselection unlikely.
All tool names follow a consistent verb_noun pattern with the prefix 'vaultfire_' (e.g., vaultfire_discover_agents, vaultfire_get_agent). This uniform snake_case naming convention makes the set predictable and easy to parse, with no deviations in style or structure across the seven tools.
With 7 tools, the count is well-scoped for a server focused on AI agent verification and data retrieval in the Vaultfire ecosystem. Each tool serves a specific, non-redundant function, covering discovery, identity, bonds, reputation, scoring, stats, and verification, which aligns perfectly with the domain's needs without being overly sparse or bloated.
The tool surface provides complete coverage for the domain of AI agent verification and on-chain data access. It includes discovery, detailed identity and bond retrieval, reputation and scoring metrics, protocol-level statistics, and a comprehensive verification tool that integrates all aspects. There are no obvious gaps; agents can perform end-to-end workflows without dead ends.
Related MCP Connectors
Trust infrastructure for AI agents. Portable reputation (JTS 0-5), agent discovery, vouching.
Trust infrastructure for AI agents: read a verifiable Trust Rating, claim an identity, earn a badge.
Verified EVM, Aptos and finance tools for AI agents; gas for 14 chains. Pay per call (x402).
AI Agent Trust Stack MCP: Veritas Proof-Chain Notary, Machine Credit Bureau, Firewall & Grader.