智能代理控制 是 AI Skill Hub 本期精选MCP工具之一。综合评分 8.0 分，整体质量较高。我们强烈推荐将其纳入你的 AI 工具库，帮助提升工作效率。
智能代理控制 是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。
智能代理控制 是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。
# 方式一：通过 Claude Code CLI 一键安装
claude skill install https://github.com/agentveil-protocol/agentveil-sdk
# 方式二：手动配置 claude_desktop_config.json
{
"mcpServers": {
"------": {
"command": "npx",
"args": ["-y", "agentveil-sdk"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
# 安装后在 Claude 对话中直接使用 # 示例： 用户: 请帮我用 智能代理控制 执行以下任务... Claude: [自动调用 智能代理控制 MCP 工具处理请求] # 查看可用工具列表 # 在 Claude 中输入："列出所有可用的 MCP 工具"
// claude_desktop_config.json 配置示例
{
"mcpServers": {
"______": {
"command": "npx",
"args": ["-y", "agentveil-sdk"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
<img src="docs/logo.png" alt="AgentVeil" width="160">
Action control surface
eddsa-jcs-2022 Data Integrity proofsSupporting signals (advisory)
---
Project owners can use AgentVeil as an action-control path for agents, tools, workflows, MCP servers, and CI jobs inside one project.
pip install lurkr
lurkr scan --path ./your-agent-project
agentveil policy init # (planned for v0.8 / Phase 3)
from agentveil import evaluate_action # (planned for v0.8 / Phase 3)
4. Produce signed evidence today with controlled_action(...), DelegationReceipts, approval routing, and Proof Packets.
See Mode A Quickstart for the full Project Owner path and planned capability markers.
---
| Example | Description |
|---|---|
[first_controlled_action.py](examples/first_controlled_action.py) | **Action control demo** — preflight → Runtime Gate → approval routing → signed receipt |
[approval_flow.py](examples/approval_flow.py) | **Approval pattern** — controlled_action → approval_required → grant → execute_after_approval |
[handle_errors.py](examples/handle_errors.py) | **Error patterns** — typed exception handling for retry, re-auth, validation, network |
[proof_packet_export.py](examples/proof_packet_export.py) | **Proof packet export** — build, save, reload, verify offline (mock mode) |
[registration/](examples/registration/) | **Registration patterns** — first-time setup, verification state, encrypted reload |
[delegation/](examples/delegation/) | **DelegationReceipt patterns** — issue, verify offline, persist/reload, multi-scope |
[proof_pack/](examples/proof_pack/) | **Offline audit verification** — local-backend demo: signed events → tamper-resistant chain → independent offline verification (no SDK or AVP API needed). Local backend required. |
[standalone_demo.py](examples/standalone_demo.py) | **Agent network primitives** — registration, peer attestations, scoring (mock mode, no server). Advanced internal surface. For action control, see [Mode A Quickstart](docs/MODE_A_QUICKSTART.md). |
[quickstart.py](examples/quickstart.py) | Register, publish card, check reputation |
[two_agents.py](examples/two_agents.py) | Full A2A interaction with attestations |
[verify_credential_standalone.py](examples/verify_credential_standalone.py) | Offline credential verification (no SDK needed) |
Framework examples: CrewAI · LangGraph · AutoGen · OpenAI · Claude MCP · Paperclip
---
```bash
cred = agent.get_reputation_credential(format="w3c") assert AVPAgent.verify_w3c_credential(cred) # offline, no API call ```
---
from agentveil import AVPAgent
agent = AVPAgent.load("https://agentveil.dev", "my-agent")
report = agent.integration_preflight()
if not report.ready:
raise RuntimeError(report.next_action)
outcome = agent.controlled_action(
action="deploy.release",
resource="service:critical-workflow",
environment="production",
delegation_receipt=delegation_receipt, # issued by the workflow owner
)
if outcome.status == "approval_required":
wait_for_principal_approval(outcome.approval["approval_id"])
elif outcome.status == "executed":
store(outcome.receipt_jcs)
elif outcome.status == "blocked":
raise RuntimeError(outcome.reason)| Stack | Install | Integration surface |
|---|---|---|
| **Any Python** | pip install agentveil | AVPAgent, integration_preflight(), controlled_action(), build_proof_packet() |
| **CrewAI** | pip install agentveil crewai | AVPReputationTool, AVPDelegationTool, AVPAttestationTool |
| **LangGraph** | pip install agentveil langgraph | ToolNode([avp_check_reputation, avp_should_delegate, avp_log_interaction]) |
| **AutoGen** | pip install agentveil autogen-core | avp_reputation_tools() |
| **OpenAI** | pip install agentveil openai | avp_tool_definitions() + handle_avp_tool_call(...) from agentveil.tools.openai |
| **MCP clients** | pip install 'agentveil[mcp]' | agentveil-mcp toolbox for explicit Runtime Gate evaluation, approvals, receipts, reputation, identity lookup, and audit. It does not intercept or gate other MCP tools. ([docs](agentveil_mcp/README.md)) |
| **MCP transport proxy** | pip install agentveil | agentveil-mcp-proxy wraps downstream MCP servers (filesystem, github, shell) with Action Control Plane gating, approval routing, durable signed evidence, and replay defense for Claude Desktop, Cursor, Cline, Windsurf, and VS Code ([docs](agentveil_mcp_proxy/README.md)) |
| **Gemini** | pip install agentveil google-generativeai | Function-calling example: [examples/gemini_example.py](examples/gemini_example.py) |
| **PydanticAI** | pip install agentveil pydantic-ai | Tool example: [examples/pydantic_ai_example.py](examples/pydantic_ai_example.py) |
| **Paperclip** | pip install agentveil | avp_should_delegate(...), avp_evaluate_team(...), avp_plugin_tools() |
| **AWS Bedrock** | pip install agentveil boto3 | Converse API example: [examples/aws_bedrock.py](examples/aws_bedrock.py) |
| **Microsoft AGT / AgentMesh** | pip install agentmesh-avp | agentmesh-avp integration package for Agent Governance Toolkit / AgentMesh |
Full integration guides: docs/INTEGRATIONS.md
---
| Without AgentVeil | With AgentVeil | |
|---|---|---|
| **Risky capability discovery** | Found in incident review | Pre-runtime posture check finds bypass paths, exposed credentials, missing approvals |
| **Risky action execution** | Agent calls deploy / transfer / delete directly | Evaluated before execution → allow / approval_required / block |
| **Approval on critical steps** | Rubber-stamped or skipped | Signed approval receipt — single-use, expiring, bound to exact action/resource/env |
| **Audit evidence** | "Agent triggered X" in app logs | Signed receipt with action hash, decision hash, approval hash, timestamp — verifiable offline by audit / customer / partner |
---
高质量的开源MCP工具，具有较强的实用价值
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一，可自由商用、修改、分发，仅需保留版权声明。
经综合评估，智能代理控制 在MCP工具赛道中表现稳健，质量优秀。如果你已有明确的使用需求，可以直接上手体验；如果还在评估阶段，建议对比同类工具后再做决策。
| 原始名称 | agentveil-sdk |
| 原始描述 | 开源MCP工具：Action Control for autonomous agents. Check posture. Gate risky actions. Prove e。⭐9 · Python |
| Topics | agent-controlai-agentscontrolled-actions |
| GitHub | https://github.com/agentveil-protocol/agentveil-sdk |
| License | MIT |
| 语言 | Python |
收录时间：2026-05-26 · 更新时间：2026-05-26 · License：MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型，复制安装指令后粘贴到对应客户端