OwlCore.Storage.Mcp

# OwlCore.Storage.Mcp A Model Context Protocol (MCP) server that provides unified storage operations across multiple protocols including local files, IPFS MFS, HTTP/HTTPS, and custom mounted folders. This project bridges the gap between different storage systems by providing a consistent MCP interface for file operations. Whether you're working with local files, IPFS content, web resources, or custom storage protocols, OwlCore.Storage.Mcp offers a unified API that abstracts away the underlying complexity. The plugin-style architecture makes it easy to extend support for new storage protocols while maintaining backward compatibility. ## Featuring: - **Multi-protocol support**: Local files, IPFS MFS, HTTP/HTTPS, IPFS/IPNS content, memory storage - **Custom mount system**: Mount any folder with a custom protocol scheme for organized access - **Persistent mounts**: Mount configurations are saved and restored across sessions - **Plugin architecture**: Easy to extend with new storage protocols via `IProtocolHandler` - **Performance optimized**: Storage registry pattern prevents object recreation - **Cross-platform**: Built on .NET 9.0 for maximum compatibility ## Usage ### Setup 1. **Clone and build the project**: ```bash git clone https://github.com/Arlodotexe/OwlCore.Storage.Mcp.git cd OwlCore.Storage.Mcp dotnet build ``` 2. **Configure with MCP clients**: - The server communicates via stdio (no console output visible) - Configured and launched automatically by MCP clients - Server provides unified storage tools accessible through MCP client interfaces ### Client Configuration This MCP server is designed to be used locally only, over stdio. For **LM Studio** ([setup guide](https://lmstudio.ai/docs/app/plugins/mcp#install-new-servers-mcpjson)), add to `mcpServers`: ```json { "mcpServers": { // Add the "OwlCore.Storage" server config object here } } ``` For **GitHub Copilot in VS Code** ([setup guide](https://code.visualstudio.com/docs/copilot/chat/mcp-servers)), add to `servers`: ```json { "servers": { // Add the "OwlCore.Storage" server config object here } } ``` In both cases, you'll need this server config object: ```json "owlcore.storage": { "command": "dotnet", "args": ["run", "--project", "/path/to/OwlCore.Storage.Mcp/"] } ``` #### Recommended system prompt Not all language models or agent software comes with a usable tool-calling system prompt out of the box. This system prompt was tested extensively on `mistralai/devstral-small-2507` and is free to use and modify: ``` You are a helpful and capable AI assistant who is able to call tools to assist the user, an assumption-less skeptic even when confident, and you always think carefully before you act. TOOL GUIDELINES: - Only call the tools that have been explicitly declared as available to you, do not call **any** other tools, EVER. - Never say the name of a tool to a user-- Instead, state the intent regarding the tool. - Do not tool call unless asked or implied by the user. - Use the least tool calls reasonably possible to complete the request, and use as many as is necessary to complete the request autonomously. - Keep Unicode escaped, never unescape Unicode. ``` ### Available Storage Operations Once connected, you can use MCP tools to: - Browse local files, IPFS MFS, and HTTP resources - Mount custom folders with personalized protocol schemes - Read/write files across different storage systems - Navigate folder hierarchies with unified commands ## Financing We accept donations [here](https://github.com/sponsors/Arlodotexe) and [here](https://www.patreon.com/arlodotexe), and we do not have any active bug bounties. ## Versioning Version numbering follows the Semantic versioning approach. However, if the major version is `0`, the code is considered alpha and breaking changes may occur as a minor update. ## License All OwlCore code is licensed under the MIT License. OwlCore is licensed under the MIT License. See the [LICENSE](./src/LICENSE.txt) file for more details.