Tiger Skills MCP Server emulates Claude Skills within an MCP-compatible agent by loading a curated set of modular Skills and enabling a subagent workflow for breaking down complex tasks. It exposes an environment where you can enable or disable Skills, control tool usage, and load domain-specific knowledge to guide the agent's reasoning and actions. The server supports a Subagent Task Execution flow, allowing tasks to be decomposed into subtasks each handled by separate agent instances while sharing the same configured Skill set. You can instantiate the server, connect a client, and request tasks that leverage the included Skills to perform domain-specific procedures, workflows, and tool integrations. Tools and resources can be toggled on or off via the connection parameters, giving you a resource-efficient mode for pure reasoning or a full-featured mode with tools and references loaded into context when needed.

Prerequisites:

• Python 3.9+ (or a compatible Python runtime)
• git installed on your system
• Optional: pipx for isolated tool installation

Install steps:

1. Clone the repository (or install from your package index if published): git clone https://github.com/timescale/tiger-skills-mcp-server.git cd tiger-skills-mcp-server

2. Set up a Python environment (recommended): python3 -m venv venv source venv/bin/activate

3. Install the MCP server package (via uvx, per this project’s configuration):

   If uvx is available in your environment

   pip install uvx

4. Install the Tiger Skills MCP server package (assuming it is published as a package named timescale-tiger-skills-mcp-server): uvx install timescale-tiger-skills-mcp-server

   or, if using a local repo: python -m pip install -e .

5. Configure environment variables (example): export SKILLS_FILE=./skills.yaml export SKILLS_TTL=300000

6. Run the server (example using uvx as configured): uvx run timescale-tiger-skills-mcp-server

Note: If you prefer a containerized approach and a prebuilt image is available, you can adapt these steps to a docker workflow (see the Docker section in the project documentation if provided).

Tips and considerations:

• Ensure SKILLS_FILE points to a valid YAML configuration describing your skill set. The server caches skills according to SKILLS_TTL (in milliseconds).
• The Subagent feature enables task decomposition. Use it when tasks are large or data-heavy to manage context efficiently.
• Connection string parameters can be used to enable/disable skills and tools, for example: ?enabled_skills=a,b&tools=1&resources=0.
• If you experience issues loading skills, verify YAML syntax in the skills file and ensure there are no duplicate skill names across configured collections.
• The server URL pattern in readme examples suggests an HTTP interface at /mcp for client connections; adjust base URLs if you deploy behind reverse proxies or in a containerized environment.