AgentSeal
agentseal-mcp
AgentSeal 的 MCP 服务器。使用 SHA-256 哈希链为 AI Agent 提供防篡改审计日志。
每个 Agent 的操作都记录在哈希链中。有了它,您可以向客户证明您的 Agent 确实执行了其声称的操作。
设置
Claude Desktop
添加到您的 claude_desktop_config.json:
{
"mcpServers": {
"agentseal": {
"command": "npx",
"args": ["-y", "agentseal-mcp"],
"env": {
"AGENTSEAL_API_KEY": "as_sk_your_key_here"
}
}
}
}保存后重启 Claude Desktop。
Cursor / 其他 MCP 主机
配置相同 — 使用您的 API 密钥添加服务器。
环境变量
变量 | 必需 | 描述 |
| 是 | 您从 agentseal.io 获取的 API 密钥 |
| 否 | 自定义 API 基础 URL(默认为生产环境) |
Related MCP server: Agent Audit Trail MCP Server
工具
record_action
将 Agent 操作记录到审计追踪中。在执行重要操作后调用此工具,以创建关于操作内容和原因的加密链式记录。
参数 | 类型 | 必需 | 描述 |
| string | 是 | Agent 的标识符(例如 |
| string | 是 | 操作类型(例如 |
| object | 否 | 操作详情 |
| string | 否 | Agent 决定执行此操作的原因 |
| string | 否 | 谁或什么批准了该操作 |
返回序列号和 SHA-256 哈希值,确认条目已链接。
query_actions
从审计追踪中查找之前记录的操作。使用此工具检查已执行的操作或回顾过去的决策。
参数 | 类型 | 必需 | 描述 |
| string | 否 | 按 Agent 筛选 |
| string | 否 | 按操作类型筛选 |
| number | 否 | 返回的最大条目数(默认为 20) |
verify_chain
验证哈希链的完整性。每个条目的 SHA-256 哈希都包含前一个条目的哈希值——如果任何记录被修改,链条就会断裂,此工具会报告断裂位置。
参数 | 类型 | 必需 | 描述 |
| string | 否 | 验证特定 Agent 的链。如果省略,则验证所有条目。 |
返回已验证的条目数量以及链条是否完整。
工作原理
每个记录的操作都使用 SHA-256 进行哈希处理。该哈希值包含前一个条目的哈希值,从而形成一条链。修改任何记录,其后的所有哈希值都会改变——verify_chain 会立即检测到。
获取 API 密钥
在 agentseal.io 注册。免费使用。
Python SDK
如需在不使用 MCP 的情况下直接集成:pip install agentseal-sdk。请参阅 agentseal-sdk。
许可证
MIT
Available Tools
3 toolsquery_actionsA
Look up previously recorded actions from the audit trail. Use this to check what actions have been taken, verify history, or recall past decisions.
Inputs: agent_id, action_type, limit.
record_actionA
Record an agent action to the AgentSeal audit hash trail. Call this after every significant action (sending emails, modifying files, running queries, making API calls) to create a cryptographically sealed record of what happened and why.
Inputs: agent_id (required), action_type (required), action_params, reasoning, +1 more.
verify_chainA
Verify the integrity of the audit trail hash chain. Each entry's SHA-256 hash includes the previous entry's hash — if any record was modified, the chain breaks and this will report where.
Inputs: agent_id.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: record_action records, query_actions retrieves, verify_chain checks integrity. No overlap or ambiguity.
All tool names follow a consistent verb_noun pattern (query_actions, record_action, verify_chain), making them predictable and easy to understand.
Three tools is an ideal scope for an audit trail server, covering the essential operations of recording, querying, and verifying without unnecessary bloat.
The tool surface fully covers the core lifecycle of an audit trail: recording actions, looking up history, and verifying chain integrity. No obvious gaps given the domain's immutability requirements.
Maintenance
Related MCP Connectors
Signed, publicly logged records of what your AI agent did, with proofs anyone can check offline.
Bitcoin-anchored, tamper-evident audit log for AI agents — record, disclose and verify actions.
Issue & verify signed (ed25519), hash-chained, timestamped provenance receipts for agent actions.
Immutable event logging and audit trail for agent transactions
Related MCP Servers
AlicenseAqualityCmaintenanceA Model Context Protocol server that enables LLMs to retrieve and analyze OpenTelemetry traces and metrics from Logfire, supporting exception tracking and custom SQL queries against telemetry data.435,280 PyPI162MIT- AlicenseNot gradedqualityDmaintenanceProvides tamper-proof audit logging for AI agents using SHA-256 hash chains, integrity verification, and compliance reporting for the EU AI Act.27 PyPI1MIT
- AlicenseNot gradedqualityDmaintenanceProvides an immutable, tamper-evident audit trail for AI agents, enabling event logging with cryptographic chaining, search, verification, and statistics.3MIT
- AlicenseNot gradedqualityBmaintenanceTamper-evident audit trail for AI agent tool calls. Cryptographically signed, hash-chained records that prove what your agents did.3MIT