Praxia
Praxia 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。
📚 深度解析
通过安装 Praxia，你的 AI 助手将获得额外的工具调用能力，可以用自然语言直接操控该工具的功能，无需学习复杂的命令行语法。MCP 工具的核心价值在于"一次配置，永久增强"——配置完成后，每次与 AI 对话时都可以无缝调用这些工具。
在技术实现上，MCP 工具通过标准的 JSON-RPC 协议与 AI 客户端通信，工具的功能以"工具列表"的形式暴露给 AI 模型，AI 可以按需调用。Praxia 提供了结构化的工具调用接口，使 AI 模型能够精确地理解和使用每个功能点，显著降低 AI 在工具使用上的错误率。
与传统的 API 集成相比，MCP 工具的优势在于无需编写代码——用户只需在配置文件中添加几行 JSON，即可让 AI 获得全新能力。AI Skill Hub 将 Praxia 评为 AI 评分 8.0 分，属于同类工具中的优质选择。
📋 工具概览
Praxia 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。
📖 中文文档
Praxia 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。
- 通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
- 提供结构化工具调用接口，显著降低 AI 集成复杂度
- 支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
- 可与其他 MCP 工具组合叠加，构建完整 AI 工作站
- 轻量无侵入设计，不影响现有系统架构
- 在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
- 通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
- 将多个 MCP 工具组合使用，构建个人专属 AI 工作站
# 方式一：通过 Claude Code CLI 一键安装
claude skill install https://github.com/praxia-dev/praxia
# 方式二：手动配置 claude_desktop_config.json
{
"mcpServers": {
"praxia": {
"command": "npx",
"args": ["-y", "praxia"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
- 确认已安装 Node.js（v18 或以上版本）
- 打开 Claude Desktop 或 Claude Code 的 MCP 配置文件
- 按「交给 Agent 安装 → Claude Desktop」标签中的 JSON 配置填入 mcpServers 字段
- 保存配置文件并重启 Claude 客户端
- 重启后，在对话中即可使用本工具
# 安装后在 Claude 对话中直接使用 # 示例： 用户: 请帮我用 Praxia 执行以下任务... Claude: [自动调用 Praxia 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": {
"praxia": {
"command": "npx",
"args": ["-y", "praxia"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
Praxia
🌐 Live: praxia.tools (primary, Cloudflare) · praxia-dev.github.io/praxia (mirror, GitHub Pages) · @praxia_dev on X
<sub>📺 Watch the 4-minute walkthrough · 🚀 Quickstart · 💬 Discussions</sub>
✨ What's Bundled
What's new in alpha22–27
alpha22 — Documents semantic search. Every chunk uploaded to a Documents folder is embedded via litellm (auto-detected from your provider env vars: OpenAI →text-embedding-3-small, Ollama →nomic-embed-text, Azure / Gemini / DashScope / HuggingFace also covered). Cosine-similarity ranking replaces keyword scoring, so a JA query like "アクションアイテム" finds EN chunks about action items. Pre-alpha22 docs back-fill silently on first boot after upgrade. alpha22 — Batch fan-out via natural language. Say "Documents の各 PDF からアクションアイテムを抽出して" / "for each PDF in folder X, summarise" and the agent classifier routes to abatchtask kind that lists files, fans out one prompt per file viarun_parallel_tasks, and surfaces composite progress in the Batches tab. alpha22 — OSS license disclosure. Settings → "Open-source notices" shows the full text of every third-party license shipped in the desktop installer (562 packages: 71 Python · 469 Rust · 22 npm). The full text is bundled with the installer for compliance with MIT / BSD / Apache 2.0 / MPL. alpha23 — Auth propagation across worker tasks. Batches and Schedules now carry the parent agent's LLM (model + scout + workspace + org) into every child TaskRecord. Previously fan-outs from non-Anthropic users crashed withMissing ANTHROPIC_API_KEYbecause the worker fell back tomodel="claude". alpha25 —web_searchagent tool. Tavily + Brave Search backends. The agent uses this for current events / today's prices / breaking news / recent regulatory changes — anything not in the user's Documents or memory layers. Tavily wins when both keys are set (returns a one-paragraph synthesised answer). No key configured → tool returns a structured error and the LLM relays it. Only the query string leaves the machine — Documents and memory content stay local. alpha26 — Promotion gate + bitemporal SharedMemory.PromotionEnginenow takes averifiercallback that short-circuits the score blend to"skip"onpassed=False, plusmin_independent_sources=2(single-source candidates cap at "review" instead of auto-promoting).SharedBlockis bitemporal- lite:upsert()supersedes the old block (valid_to=now,superseded_by=<id>) rather than overwriting;as_of(timestamp)queries past snapshots.search()tie-breaks ties by verification_score → source_breadth → recency. Built from the L0–L3 verification findings — see release notes for details. alpha27 — LLM-driven query expansion (Anthropic fallback). When no embedding provider is configured (most common: Anthropic- only, since Claude has no embedding API), Documents search no longer falls back to pure literal keyword matching. The active LLM rewrites the query into 3–5 alternative phrasings (synonyms, cross-language equivalents, paraphrases) and each chunk is scored against the max-scoring variant. Results cached LRU per(model, query). Anthropic-only users now get usable cross-language Documents search without a second provider key — at the cost of one small extra LLM round-trip per distinct query.
Auto-dispatched by extension:
| Extension | Parser | Optional dep |
|---|---|---|
.txt .md .rst .py .ts .js | TextParser | (none) |
.csv .tsv | CsvParser | (stdlib) |
.json .yaml .yml | StructuredParser | (core) |
.html .xml | HtmlParser | (stdlib) |
.pdf | PdfParser | praxia[office] |
.docx | DocxParser | praxia[office] |
.pptx | PptxParser | praxia[office] |
.xlsx .xlsm | XlsxParser | praxia[office] |
from praxia.io.parsers import parse_file
doc = parse_file("contract.pdf") # works
doc = parse_file("Q3_results.xlsx") # also works
print(doc.content)Third-party formats register via [project.entry-points."praxia.parsers"] — no fork required.
2. Promote only what's **effective**, **automatically**
Three independent verdicts run in parallel. The framework auto-promotes only when consensus is high; medium-confidence items go to a review queue.
Skipped by default (requires API keys + costs tokens)
pytest tests/llm_eval -m llm_eval -v
Install Praxia Desktop (Windows 10 / 11 x64)
Specialized Multi-Agent Orchestrator with Cyclic Personal/Organizational Memory A workflow-specific multi-agent orchestrator that automatically promotes individual tacit knowledge into organizational know-how. Built on a 5-layer memory stack with three independent promotion paths.
[
]() [
]() [
]() [
]() [
]()
🔍 Complete feature reference: docs/FEATURES.md 📊 Concrete Before/After tables: docs/use-cases.md
---
⬇ **Try Praxia in 30 seconds — no Python, no setup**
1. Install (pick the extras you actually need)
pip install praxia # Core pip install "praxia[ui,connectors,office,audio]" # Common stack pip install "praxia[all]" # Everything
🚢 Deployment modes
Praxia ships in two halves you can mix:
| Mode | What you run | When to choose it |
|---|---|---|
| **A. Full-stack** | praxia ui (Streamlit) + Praxia core, one process | Internal team, fastest path |
| **B-1. Embedded SDK** | Your Python service import praxia | You already have a Python backend |
| **B-2. HTTP service** | praxia serve (FastAPI) + your own frontend | Non-Python frontend, mobile, or CDN-cached UI |
Both modes share the same auth, memory, and skills — only the frontend differs. Step-by-step setup, production checklist, and migration path: docs/deployment-modes.md (JA).
```bash
🚀 Quickstart
### 🖥 Native desktop app (no Python required) The easiest way to try Praxia is the native desktop app on the Microsoft Store. | Platform | Distribution | Status | |---|---|---| | Windows 10 / 11 x64 | Microsoft Store | ✅ shipped | | macOS 12+ | — | 🚧 Phase 1b | | Linux (Debian / Ubuntu) | — | 🚧 Phase 1b | 👉 Get Praxia Desktop on the Microsoft Store — installed through the Store, so no SmartScreen warning and updates land automatically. Past release notes remain on GitHub Releases. The desktop app embeds the Praxia server inside the installer — install, launch, paste an LLM provider key, and you're running. No separatepraxia serveprocess to start, nopip install, no Python on the user's machine. Settings exposes only the three things a user actually controls: LLM provider keys (Anthropic / OpenAI / Google / Azure OpenAI / Qwen DashScope / Hugging Face — Gemma covered via all three cloud paths), local LLM (Ollama URL + model), and optional SSO tenant URL for org connection. Everything else (port, API key, storage layout, CORS) is managed by the app. Multi-user organizational deployment still works the same way: install Praxia on a shared host aspraxia serve, point a custom frontend or SDK consumer at it, and several users share L3 organizational memory + the L4 frozen layer with SSO / RBAC / audit / KMS-encrypted OAuth tokens — all in the OSS core. Desktop-only features (in addition to everything the server offers): - 🗂 Local folder ingestion with auto-discovery — point Praxia at a folder on your machine (e.g.~/Documents/Contracts/); the desktop app walks it recursively, parses every supported file (PDF / DOCX / PPTX / XLSX / TXT / MD / code), and makes the contents searchable by the agent alongside L1 / L3 / L4 memory. Skips files above a size cap, watches for new / changed files, re-indexes incrementally by mtime + content hash. Useful for confidential documents you don't want uploaded to cloud storage. (🚧 Phase 1b) - 🔔 Native notifications when a long-running agent task finishes (🚧 Phase 1b) - 🪟 Native file dialogs for drag-and-drop attachments Cross-device continuity (PC ↔ phone handoff) and a mobile companion land in Phase 1b / Phase 2. --- For library / SDK / CLI use, the Python install below is the developer path.
```bash
Or: cp .env.example .env and edit
a fresh key. See docs/quickstart.md § 3a for the full first-login flow.
📊 Use Cases by Industry
Detailed Before/After tables for each domain are in docs/use-cases.md. Highlights:
| Industry | Representative use case | Headline impact |
|---|---|---|
| Investment | Seed-stage VC due diligence | 4–6h → **45–60 min** per deck |
| Sales | Pre-meeting research + storyboard | Proposal-acceptance rate **+15–20pt** |
| Engineering Design | Requirements doc review | Senior architect time freed: **week 16h → 4h** |
| Procurement | RFQ TCO comparison | Hidden costs found: **+30%** vs initial quote |
| Patent | Prior-art search + novelty assessment | External patent-attorney fees **−50–70%** |
| Legal | M&A contract review | External law-firm costs **halved** (~$100k/deal) |
3-year compounding effects: New-hire ramp 6–12mo → 2–3mo / Veteran-departure knowledge loss → zero / Cross-team best-practice diffusion 30+ items/month.
---
2. Configure once — all keys live in one place
praxia config init # interactive walkthrough praxia config show # display resolved config (secrets masked) praxia config path # show key resolution order
<provider>" via OIDC SSO when PRAXIA_SSO_PROVIDER is configured.
Each user authorizes individually (CLI loopback)
praxia oauth start box --user-id alice
↑ also bootstraps an `admin` user and writes its API key (one-time only)
to `.praxia/auth/BOOTSTRAP_API_KEY.txt`. Save it somewhere safe and
Dashboard / Preferences (+ Admin for admin role). Login: User ID +
Password (= API key from `praxia user create`), or "Sign in with
Backend-only HTTP API (8 endpoints under /api/v1)
pip install "praxia[server]" praxia serve --host 0.0.0.0 --port 8000 --cors-origin https://your-frontend.example ```
---
(b) Entry-point (third-party packages — no fork needed)
[project.entry-points."praxia.connectors"] notion = "praxia_connector_notion:NotionConnector" ```
After pip install praxia-connector-notion, the new connector shows up automatically in praxia connector list, the Streamlit UI, and the SDK — with no edit to Praxia itself.
Full guide with examples for all 4 plugin types: docs/PLUGINS.md.
---
Compare providers on the same cases
pytest tests/llm_eval --llm-eval-model gpt-4o ```
Built-in rubrics: keyword match, structure (heading) match, length band, must-not-contain, LLM-as-judge. One canonical case per business skill ships out of the box.
Praxia 是一个用于自然语言处理的开源框架，提供了一个易于使用的 API，支持多种语言模型和服务。它旨在简化开发过程，提供高效的自然语言处理功能。
Praxia 包括以下功能： • 自动推送高效的、自动推送的内容 • 三个独立的判定结果并行运行 • 只有当共识很高时才会自动推送 • 中等可信度的项目会进入审查队列 • 支持多种语言模型和服务
Praxia 需要 API 密钥和成本 token，需要 pytest 进行测试
Praxia 可以通过以下方式安装: • 下载 Praxia 桌面应用 (.exe, 165 MB) • 使用 pip 安装 Praxia • 使用 Docker 安装 Praxia • 从源码安装 Praxia
使用 Praxia 的步骤如下: • 下载 Praxia 桌面应用 (.exe, 165 MB) • 使用 pip 安装 Praxia • 使用 Docker 安装 Praxia • 从源码安装 Praxia • 配置 Praxia • 使用 Praxia
Praxia 的配置包括: • 使用 MCP 配置 Praxia • 使用环境变量配置 Praxia • 配置关键参数
Praxia 的 API 包括: • OAuth API • CLI API • SDK API
Praxia 的工作流包括: • 使用 Praxia 的步骤 • 配置 Praxia • 使用 Praxia • 使用 Praxia 的 API
Praxia是一个高质量的开源MCP工具，具有较强的实用价值
- 使用 Cursor 编辑器、希望提升 AI 编程效率的开发者
- 需要让 Claude / Cursor 操作本地工具的 AI 工程师
- 构建多智能体协作系统的 Agent 开发者
- 构建企业知识库 / RAG 检索应用的团队
- 跨境业务、多语言内容运营团队
- 做语音类 AI 产品的开发者
- 配置 MCP 服务器时建议使用 stdio 传输 + JSON-RPC，避免暴露公网
- 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
- 分块大小建议 256-512 tokens，向量库优选 pgvector 或 Qdrant
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- Cursor rules 控制在 80 行内，否则模型上下文成本会显著上升
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- MCP 配置路径拼错或权限不足，重启 Claude Desktop 才生效
- embedding 模型与查询模型不一致导致检索失效
- 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
- Python 依赖冲突：建议用 venv / uv 隔离环境
- CLI：直接 npm install -g / pip install，命令行调用
- 本地部署：CPU 8GB 起，GPU 推荐 16GB+ 显存
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
- 提供结构化工具调用接口，显著降低 AI 集成复杂度
- 支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
- 可与其他 MCP 工具组合叠加，构建完整 AI 工作站
- 轻量无侵入设计，不影响现有系统架构
- 使用 Cursor 编辑器、希望提升 AI 编程效率的开发者
- 需要让 Claude / Cursor 操作本地工具的 AI 工程师
- 构建多智能体协作系统的 Agent 开发者
- 构建企业知识库 / RAG 检索应用的团队
- 配置 MCP 服务器时建议使用 stdio 传输 + JSON-RPC，避免暴露公网
- 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
- 分块大小建议 256-512 tokens，向量库优选 pgvector 或 Qdrant
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- MCP 配置路径拼错或权限不足，重启 Claude Desktop 才生效
- embedding 模型与查询模型不一致导致检索失效
- 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
👥 适合人群
🎯 使用场景
- 在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
- 通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
- 将多个 MCP 工具组合使用，构建个人专属 AI 工作站
⚖️ 优点与不足
- +Apache-2.0 协议，可免费商用
- +标准化 MCP 协议，生态互联性强
- +与 Claude 官方生态无缝对接
- +即插即用，配置简单快捷
- −依赖 Claude 客户端，非 Claude 用户无法使用
- −MCP 协议仍在持续演进，接口可能变更
- −需要一定的配置步骤
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。
✅ Apache 2.0 — 宽松开源协议，可商用，需保留版权声明和 NOTICE 文件，含专利授权条款。
🔗 相关工具推荐
❓ 常见问题 FAQ
经综合评估，Praxia 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 | praxia |
| 原始描述 | 开源MCP工具：Multi-agent orchestrator with cyclic personal-to-org memory · Apache 2.0 · Pytho。⭐8 · Python |
| Topics | ai-agentsapache-2-0knowledge-management |
| GitHub | https://github.com/praxia-dev/praxia |
| License | Apache-2.0 |
| 语言 | Python |
收录时间：2026-06-07 · 更新时间：2026-06-08 · License：Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。