pokemon-chat exposes an MCP-compliant service that orchestrates a knowledge-guided Pokémon conversation assistant. The server combines a knowledge graph, a vector store, and multi-agent orchestration to provide precise, context-aware answers about Pokémon lore, moves, types, locations, and related data. With MCP support, you can connect real-world geospatial data and local knowledge sources to extend the assistant’s capabilities, enabling tasks such as querying the knowledge graph, performing semantic searches, and invoking agent-driven workflows. Use the MCP endpoints to fetch contextual responses, subtools, and task flows, and leverage the included multi-agent coordination (Supervisor + Workers) to run parallel subqueries or reasoning steps as needed.

Prerequisites:

• Docker and Docker Compose installed on your machine
• Git (optional but recommended)
• Basic familiarity with MCP concepts and environment variables

Installation steps:

1. Clone the repository (if you have access to the source): git clone https://github.com/skygazer42/pokemon-chat.git cd pokemon-chat

2. Prepare environment variables (example): cp .env.example .env

   Edit .env to enable MCP and any features you want (e.g., enable_knowledge_graph, enable_knowledge_base, enable_mcp)

3. Run the application via Docker (as suggested by the Docker Compose setup): docker run -i skygazer42/pokemon-chat

4. Access the MCP-enabled service (endpoints will be exposed according to the container's configuration):

   • API: http://localhost:5050/api/docs
   • Web UI: http://localhost:3100/

5. Optional: if you use the full docker-compose setup with infra and MCP profiles, follow the repository's Docker Compose instructions to bring up the infra services and MCP port mappings.

Note: This MCP setup integrates Neo4j, Milvus, and other components behind the scenes; ensure your environment has suitable resources and network access for these services.

Tips and common considerations:

• The MCP feature is designed to connect the Pokémon knowledge graph with external data sources. In the .env file, you can enable or disable components such as knowledge_graph (Neo4j), knowledge_base (Milvus), web_search, and MCP integration.
• If you are testing locally, ensure ports used by the UI and API are not blocked by firewalls. The defaults are typically 3100 for the web UI and 5050 for the API docs.
• For data initialization, the repo mentions Neo4j import steps when infra is enabled. If you skip infra, you may need to seed data manually according to your setup.
• When using MCP, you may pass environment variables or profiles (e.g., --profile infra --profile mcp) to Docker Compose to enable the MCP workflow and data sources.
• Check the repository’s .env.example for the exact variable names and expected values to avoid misconfigurations.