vestige

Vestige is an MCP server that runs the Vestige cognitive engine locally. It exposes a JSON-RPC style interface with 21 tools and 29 cognitive modules, enabling memory creation, searching, dreaming, and consolidation operations all within a single Rust-based binary or via its MCP-compatible tooling. Once started, you can interact with Vestige using its standard stdio JSON-RPC protocol, send commands to manage memories, run HyDE-based expansions, trigger recall, observe FSRS-6 decay, and monitor the 3D memory dashboard when the web UI is available. The 3D dashboard provides real-time visualization and WebSocket-driven events for memory creation, access, and dream/consolidation cycles, making it easy to observe how Vestige processes information as you interact with it.

To use Vestige, install the MCP server package (vestige-mcp-server) and run it. The server accepts JSON-RPC payloads over stdin/stdout, so you can script interactions, or use the included dashboards and integrations to drive cognitive tasks. The Quick Start setup also notes a REST/WebSocket-enabled dashboard on port 3927 for visualization, and the ecosystem includes multiple IDE integrations (Claude Code, Claude Desktop, Xcode, Cursor, VS Code, JetBrains, Windsurf) to connect the MCP engine to your development environment. When running, you’ll access tool endpoints to manage memories, perform queries, and observe the 3D neural graph in real time.

Tools and capabilities include: 21 MCP tools that implement operations like memory creation and search, HyDE expansion for enhanced recall queries, FSRS-6-based decay visualization, Nomic MoE embedding options, a memory dreaming/replay mechanism, a 3D memory dashboard, and a web-based event bus that broadcasts cognitive events for analytics and UI updates.

Prerequisites:

• curl and tar (for downloading prebuilt binaries)
• tar/gzip support on your system
• Node.js environment if you choose to install via npm (optional, see npm_package below)
• Rust toolchain if you prefer building from source (Rust 1.91+)

Install via prebuilt binaries (recommended for quick start):

1. macOS Apple Silicon (ARM): curl -L https://github.com/samvallad33/vestige/releases/latest/download/vestige-mcp-aarch64-apple-darwin.tar.gz | tar -xz sudo mv vestige-mcp vestige vestige-restore /usr/local/bin/

2. macOS Intel (x86_64): curl -L https://github.com/samvallad33/vestige/releases/latest/download/vestige-mcp-x86_64-apple-darwin.tar.gz | tar -xz sudo mv vestige-mcp vestige vestige-restore /usr/local/bin/

3. Linux (x86_64): curl -L https://github.com/samvallad33/vestige/releases/latest/download/vestige-mcp-x86_64-unknown-linux-gnu.tar.gz | tar -xz sudo mv vestige-mcp vestige vestige-restore /usr/local/bin/

4. Windows: Download from Releases page on GitHub and run the executable from the downloaded folder.

Install via npm (alternative, provides a JavaScript/Node-friendly entry):

1. Ensure Node.js and npm are installed.
2. Install the MCP server globally: npm install -g vestige-mcp-server
3. Run the server (the MCP tooling will be available via npx as configured): npx -y vestige-mcp-server

Build from source (requires Rust 1.91+):

1. git clone https://github.com/samvallad33/vestige && cd vestige
2. cargo build --release -p vestige-mcp
3. Optional: enable Metal GPU acceleration on Apple Silicon cargo build --release -p vestige-mcp --features metal

Notes and tips:

• Vestige runs entirely locally; no data leaves your machine by default, aligning with its privacy-focused design.
• The 3D memory dashboard is available at http://localhost:3927/dashboard when the MCP server starts.
• If you’re using npm to install vestige-mcp-server, you can integrate it with your existing toolchains or CI workflows via npx, npm scripts, or direct binary invocation.
• Common issues include environment path problems after installing prebuilt binaries or permission prompts when moving binaries into /usr/local/bin. Ensure your PATH includes the directory where vestige-mcp is installed.
• The MCP API is JSON-RPC over stdio or a configured interface; check docs/configuration for details on endpoints and payload formats. There are multiple integrations across IDEs and editors to connect and drive cognitive tasks and queries.
• For hardware-accelerated builds, especially on Apple Silicon, enable Metal in the build features as documented in the build steps.