A comprehensive Python framework for building, scaling, and deploying Model Context Protocol (MCP) servers and agentic applications.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install concierge-sdk
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install concierge-sdk using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Concierge SDK?
Concierge SDK is the industry-standard framework designed for developers building agentic applications using the Model Context Protocol. It simplifies the process of creating MCP servers, exposing Python functions as tools to AI agents, and connecting agents to complex APIs. By providing a structured way to handle tool disclosure and session state, it serves as a critical foundation for developers creating sophisticated Openclaw Skills.
This framework is drop-in compatible with FastMCP and raw MCP server instances, allowing developers to wrap existing codebases and immediately gain access to advanced features like semantic search, staged workflows, and cloud deployment. It is built to handle everything from simple local utilities to complex, multi-step production environments.
Concierge SDK Use Cases
- Creating custom MCP servers to expose proprietary internal APIs to AI assistants.
- Managing high-scale Openclaw Skills with hundreds of tools using semantic search to prevent token bloat.
- Implementing stateful, multi-step agentic workflows for ecommerce, data entry, or research tasks.
- Building rich UI widgets for ChatGPT using the built-in Apps SDK support.
- Transitioning local developer tools into production-ready cloud deployments.
How Concierge SDK Works
- Define tools using Python decorators that automatically generate the required JSON schemas for AI models.
- Configure stages and transitions to group tools into logical steps, ensuring the agent only sees relevant functions at the right time.
- Utilize the internal state management system to persist data between tool calls without stuffing the conversation history.
- Initialize the server using the Concierge class, which can wrap existing FastMCP or raw MCP servers.
- Run the application via stdio for local IDE integration or as a streamable HTTP app for web-based Openclaw Skills.
Concierge SDK Setup
Install the core SDK using pip:
pip install concierge-sdk
For advanced features like semantic search or PostgreSQL state backends, use the optional extras:
pip install concierge-sdk[all] # Includes semantic search dependencies
pip install concierge-sdk[postgres] # Includes PostgreSQL drivers for production
To scaffold a new project, use the CLI:
concierge init my-new-skill
Concierge SDK Data Schema & Taxonomy
The SDK organizes data through a structured state and workflow model:
| Component | Description |
|---|---|
| State | Session-scoped key-value pairs stored in memory or a distributed backend like PostgreSQL. |
| Stages | Logical groupings of tools that define the agent's visible capabilities at any given step. |
| Transitions | A directed graph defining which stages an agent can move to from its current position. |
| Metadata | Telemetry and project IDs used for cloud analytics and monitoring of Openclaw Skills. |
Concierge SDK Advanced Features
- Staged Tool Disclosure: Enforce strict execution orders and minimize token usage by dynamically filtering available tools.
- Shared Session State: Pass data between steps in an agentic workflow without relying on the LLM's context window.
- Semantic Search: Automatically hide large toolsets behind a search interface, allowing agents to find the right tool for the task dynamically.
- Streamable HTTP: Built-in support for Starlette-based HTTP servers with full CORS configuration for browser-based clients.
- Widget Support: Render HTML-based interactive UIs directly within ChatGPT conversations for more engaging Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
SSH Essentials
A comprehensive toolkit for mastering Secure Shell operations, from basic remote access to advanced port forwarding and security hardening.
ripgrep
A blazingly fast recursive search tool that respects .gitignore rules to find regex patterns across complex directories.
Linear Webhook Skill
The Linear Webhook Skill enables users to trigger specialized AI agents by simply @mentioning them within Linear issue comments.
jq JSON Processor
A lightweight and flexible command-line tool for processing, filtering, and transforming JSON data.
TuleBank
A comprehensive CLI tool for managing ARS bank transfers, crypto swaps on Base, and automated on-ramp/off-ramp workflows.
CSV to JSON Converter
A specialized utility designed to transform raw CSV data into clean, structured JSON arrays for seamless application integration.
Featured*
