AI Skill Hub 推荐使用：另号学习源機学源 是一款优质的MCP工具。AI 综合评分 7.5 分，在同类工具中表现稳健。如果你正在寻找可靠的MCP工具解决方案，这是一个值得深入了解的选择。
另号学习源機学源是一个另号学习源機学源。当前另号学习源機学源实例一个另号学习源機学源的为常子。
另号学习源機学源 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。
另号学习源機学源是一个另号学习源機学源。当前另号学习源機学源实例一个另号学习源機学源的为常子。
另号学习源機学源 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/srbarrios/agentic-test-explorer
# 方式二：手动配置 claude_desktop_config.json
{
"mcpServers": {
"--------": {
"command": "npx",
"args": ["-y", "agentic-test-explorer"]
}
}
}
# 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 另号学习源機学源 执行以下任务... Claude: [自动调用 另号学习源機学源 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", "agentic-test-explorer"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
<img src="logo.png" alt="Agentic Explorer Logo" width="500" align="center" />
A product-agnostic, AI-driven exploratory test framework that intelligently explores, tests, and validates any web application. Configure it for your stack via a small config.yaml, point it at your app, and let specialized agents drive a real browser to find bugs, render anomalies, and unscripted edge cases.
Powered by a LangGraph Swarm architecture, Playwright, and your choice of Claude (default) or Google Gemini, this framework dynamically routes tasks to behavioral QA personas and advanced stress/exploration agents, self-heals from UI errors, optionally consults user-provided MCP servers and Agent Skills for domain knowledge, generates reproducible Playwright test scripts from every bug found, and writes Markdown executive test reports.
It can also analyze GitHub Pull Requests — pass a PR URL and the framework extracts the code diff, feeds it to an LLM, and auto-generates targeted test missions covering the UI areas most likely impacted by the changes.
---
Product-Agnostic: One small config.yaml adapts the framework to any web app. Persona-Driven QA Agents: Three standard QA personas plus five advanced agents — each prompted around a specific testing strategy. Record-and-Translate Engine: Agents emit JSON intents, the deterministic engine executes and records every step to an immutable Action Tape. Every bug automatically generates a reproducible reproduction_*.spec.ts Playwright script. Resilient Selector Policy (Engine-Enforced): execute_browser_command rejects brittle XPath / positional selectors at runtime, enforcing data-test-subj → aria-label → visible text priority. Self-Healing Browser Execution: Playwright actions are wrapped to catch uncaught exceptions. Errors are returned as natural language so agents can adapt strategies. Screenshot Evidence: Agents capture full-page screenshots when bugs or anomalies are detected, then generate reproducible Playwright specs from the Action Tape. Bring-Your-Own MCP: Plug in any MCP servers via a standard mcp_servers.json — agents query them for domain knowledge instead of guessing. Bring-Your-Own Skills: Install Agent Skills (per the agentskills.io spec) under AGENT_SKILLS_ROOT and the framework exposes them automatically. Cross-Session Learning: A four-level memory system (semantic, episodic, procedural, prioritization) powered by Langmem lets agents learn across sessions. The framework remembers page structures, selector reliability, application quirks, agent observations, past bugs, and which testing strategies worked. Agent prompts and supervisor routing rules self-improve via Langmem's prompt optimizer after each batch. Agents can proactively record observations and recall past findings using semantic search. Regression Testing: Run --regression to auto-generate missions from the bug catalog — no YAML needed. The framework targets pages with known open bugs and historically flaky areas. Application Model Export: Run --export-model to export the discovered application structure (pages, selectors with reliability scores, bugs, quirks, session stats) as app_model.json. PR-Driven Test Generation: Pass a GitHub PR URL (--pr-url) and the framework extracts the diff (preferring the GitHub MCP server, falling back to gh CLI), sends it to an LLM, and auto-generates targeted mission YAML covering the UI areas impacted by the code changes. When historical bug data exists, it's injected into the LLM prompt for better-targeted missions. Optionally execute the generated missions immediately with --execute. * Automated Artifact Generation: Every test produces an isolated folder containing raw execution traces, the Action Tape, bug screenshots, reproducible .spec.ts files, and an executive Markdown report.
---
The dashboard provides four key views:
Python 3.11+ is required. A virtual environment is highly recommended.
```bash
pip install -e .
pip install -e ".[visual]"
playwright install chromium
> **Keeping dependencies up to date:** After pulling new changes, always re-sync your
> virtual environment to pick up any added or updated packages:
>
> Visual Mode requires Streamlit as an optional dependency:
```bash
pip install -e ".[visual]"
Add the --visual flag to any mission:
```bash
If you are an AI coding assistant contributing to this repository, see AGENTS.md for the conventions covering agent registration, selector policy, and tool behavior.
---
https://github.com/user-attachments/assets/9a17d846-5bee-4055-a97a-f9dd7ee191c2
---
python -m venv .venv source .venv/bin/activate
Copy .env.example → .env and fill in your values. The framework supports two LLM providers — Claude (default) and Gemini — and auto-detects which to use from available credentials.
```env
ANTHROPIC_API_KEY="your_anthropic_api_key_here"
Copy config.yaml.example → config.yaml and customize for your application:
```yaml app: name: "My Web Application" url: ${APP_URL} description: "Brief description used to give agent prompts domain context."
auth: method: form selectors: username: 'input[name="username"]' password: 'input[name="password"]' submit: 'button[type="submit"]' post_login_check: 'a[href="/home"]' # selector that confirms login worked
paths: mcp_servers: ./mcp_servers.json skills_root: ./agent-skills
llm: # provider: claude # or: gemini # claude_model: claude-sonnet-4-6 # claude_vision_model: claude-haiku-4-5 # gemini_model: gemini-3.1-flash-lite # gemini_vision_model: gemini-3.1-flash-lite
# Embedding model for semantic search in long-term memory (optional). # When configured, recall_past_findings uses vector similarity instead of # keyword matching. Gemini users can use their existing API key; Claude # users can run a local model via Ollama. # embedding_model: google-genai:models/embedding-001 # Gemini (768d) # embedding_dims: 768 # embedding_model: ollama:nomic-embed-text # Ollama local (768d) # embedding_dims: 768 ```
Copy mcp_servers.json.example → mcp_servers.json and list any MCP servers you want the agents to consult. Format follows the standard Claude Desktop / Code shape:
{
"mcpServers": {
"github": {
"transport": "http",
"url": "https://api.githubcopilot.com/mcp/"
},
"my-docs": {
"transport": "http",
"url": "https://my-docs.example.com/_mcp/"
}
}
}The github entry is used by the PR analyzer (--pr-url) to fetch PR data via MCP tools (get_pull_request, get_pull_request_diff, get_pull_request_files). If not configured, the analyzer falls back to the gh CLI.
If the file is missing or empty, agents simply run without MCP tools.
Install any Skills (per agentskills.io) under the directory pointed at by AGENT_SKILLS_ROOT (default ./agent-skills/). The framework discovers them automatically and exposes fetch_agent_skill and run_agent_skill_script to agents. If the directory is missing the framework just logs an info message.
agent-explorer --missions missions/new_user_agent.yaml
agentic-test-explorer 是一个与产品无关的 AI 驱动探索性测试框架。它能够智能地探索、测试并验证任何 Web 应用程序。开发者只需通过一个简单的 `config.yaml` 文件进行技术栈配置，即可驱动专门的 AI Agents 操作真实浏览器，自动发现 Bug、渲染异常以及各种非脚本化的边缘案例（edge cases），实现自动化的智能化测试。
本项目具备高度的通用性，通过 `config.yaml` 即可适配任何 Web 应用。核心特色在于其 Persona-Driven QA Agents 系统，内置了 3 种标准 QA 角色及 5 种高级 Agent，每种角色都针对特定的测试策略进行了 Prompt 优化。此外，框架拥有 Record-and-Translate Engine，能将 Agent 发出的 JSON 意图转化为确定性的执行步骤，并记录在不可篡改的 Action Tape 中，确保每个 Bug 都能被自动生成并追踪。
运行本项目需要 Python 3.11 或更高版本。为了保持环境整洁，强烈建议使用虚拟环境（如 venv 或 uv）进行开发。在安装依赖前，请确保您的系统已配置好相应的 Python 运行环境。
首先，建议创建并激活虚拟环境。接着，使用 `pip install -e .` 以可编辑模式安装核心包及其依赖。如果需要使用可视化 Dashboard（Streamlit），请运行 `pip install -e "[visual]"`。最后，必须通过 `playwright install chromium` 命令安装 Playwright 所需的 Chromium 浏览器内核。若后续拉取了新代码，请务必重新同步虚拟环境以更新依赖。
您可以通过命令行启动测试任务。若需启动可视化 Dashboard 实时监控测试过程，请在执行 mission 命令时添加 `--visual` 标志。例如：`agent-explorer --missions missions/new_user_agent.yaml --visual`。对于希望参与开发的 AI coding assistant，请参考 `AGENTS.md` 以了解 Agent 注册、选择器策略及工具行为的相关规范。
项目通过 `.env` 文件管理敏感信息。请将 `.env.example` 复制为 `.env` 并填写您的 API 密钥。框架支持 Claude（默认）和 Gemini 两个 LLM provider，并会根据提供的凭据自动检测。此外，通过 `config.yaml` 可以完成对目标 Web 应用的技术栈配置，实现框架与不同产品的解耦。
本项目主要通过 LLM Provider（如 Claude 或 Gemini）进行驱动。框架会自动识别环境变量中的 `GOOGLE_API_KEY` 或本地 `~/.gemini/` 配置。Agent 通过输出 JSON 格式的意图（intents）与执行引擎进行交互，确保了 AI 决策与浏览器操作之间的逻辑一致性。
本项目默认使用 3 种标准 QA 角色组成的 Swarm 进行测试，并会自动检测可用的 LLM provider（如 Claude）。如果遇到环境配置问题，请优先检查 `.env` 文件中的 API 密钥设置是否正确，以及 Playwright 浏览器是否已正确安装。
另号学习源機学源当前另号学习源機学源的为常子。当前另号学习源機学源实例一个另号学习源機学源的为常子。
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。
总体来看，另号学习源機学源 是一款质量良好的MCP工具，在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态，建议收藏备用，结合自身场景选择合适时机引入使用。
| 原始名称 | agentic-test-explorer |
| 原始描述 | 开源MCP工具：An agnostic AI-driven exploratory test framework that intelligently explores, te。⭐10 · Python |
| Topics | mcpagenticai-agentsai-testingautonomous-testingbrowser-automationpython |
| GitHub | https://github.com/srbarrios/agentic-test-explorer |
| License | MIT |
| 语言 | Python |
收录时间：2026-05-19 · 更新时间：2026-05-19 · 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