Lara Marketing Connector
A remote MCP server that provides a complete digital-marketing employee with 45 marketing skills, orchestration logic, persona, and per-client context for any MCP host.
README
title: Lara Marketing Connector emoji: 🚀 colorFrom: blue colorTo: indigo sdk: docker app_port: 7860 pinned: false license: mit
Lara Marketing Connector
A remote MCP server ("connector", AgentFactory style) that carries Lara's full marketing brain — 45 marketing skills + the "which-skill-when" orchestration logic + persona + per-client context — so any MCP host (Claude Code, Codex, OpenCode, OpenClaw, claude.ai) connects to one URL and gets the complete digital-marketing employee.
Built specs-first — see specs/. Status: v0.1 (public/demo).
Live MCP endpoint (when the Space is running):
https://rrizwan98-lara-marketing-connector.hf.space/mcp
Layout
lara-connector/
├── specs/ # 00-overview … 06-auth (read these first)
├── server.py # FastMCP gateway + all 13 tools
├── auth.py # verified identity (demo now, OAuth seam for prod)
├── session.py # HMAC session token (seals sub + tier)
├── db.py # SQLite state (users, clients, deliverables, usage)
├── config_store.py # Lara's brain: persona, rules, router map, tiers
├── skills_repo.py # serve skills/<name>/SKILL.md (path-safe)
├── skills/ # the 45 marketing skills
├── Dockerfile # Hugging Face Space (Docker SDK), port 7860
├── .github/workflows/deploy.yml # test -> (pass) -> deploy to HF
└── tests/ # test_smoke.py (in-memory) + test_http.py (live)
CI/CD
Push to main → GitHub Actions runs tests/test_smoke.py → only if tests pass, the repo is
synced to the Hugging Face Space, which rebuilds the Docker image and serves the connector.
Run locally (demo — no sign-in, all access)
pip install -r requirements.txt
python server.py # serves http://127.0.0.1:8000/mcp
python tests/test_smoke.py # 16 checks
Defaults: AUTH_DISABLED=1, GATING_ENABLED=false, PUBLIC_TIER=max, SQLite.
Set SESSION_SIGNING_SECRET before any public deployment.
The tools (13)
- Open:
health,begin_session(call first — returns persona, rules, router_map, skills, token) - config_*:
config_get_persona,config_get_rules - domain_*:
domain_list_skills,domain_get_skill,domain_route_task - user_*:
user_get_profile,user_list_clients,user_get_client_context,user_save_client_context,user_get_history,user_log_deliverable
Every tool except health/begin_session needs the session_token from begin_session.
Connect from a host
- claude.ai: Settings → Connectors → Add custom connector → paste the
/mcpURL. - Claude Code / Codex: add an MCP server pointing at the
/mcpURL. - OpenClaw: add it as an MCP server in
openclaw.json.
Tiers (free / pro / max)
Built in now, enforced later (see specs/05-tiers.md): wire OAuth
(AUTH_DISABLED=0), add billing → db.set_tier(sub, tier), set GATING_ENABLED=true.
The 4 invariants (the book)
One gateway · tools only · identity from verified sign-in (never model args) · fail closed.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。