mcp-klever-vm

This Klever MCP Server provides a knowledge base and runtime tooling geared toward Klever blockchain smart contract development. It can operate in multiple modes (HTTP API server, MCP stdio server, or a public hosted server) and supports both in-memory and Redis-backed storage for its context knowledge. The server automatically ingests and serves patterns, best practices, and examples related to Klever VM SDK usage, contract validation, and deployment workflows. You can query contexts by type or tags, retrieve specific entries, and leverage built-in modules for deployment and testing workflows. In public mode, a curated subset of read-only tools is exposed to keep the service secure while offering valuable search and retrieval capabilities. When running locally, you can use standard MCP client integrations to discover and fetch knowledge during development or CI.

Prerequisites:

• Node.js and npm/yarn/pnpm installed
• Git available to clone repositories (optional if using npx)

Installation steps:

1. Install dependencies and build (example using the repository):

# Clone the repository
git clone https://github.com/klever-io/mcp-klever-vm.git
cd mcp-klever-vm

# Install dependencies with pnpm (as used by the project)
pnpm install

# Copy environment configuration (optional for initial run)
cp .env.example .env

# Install Klever SDK tools (required for transactions)
chmod +x scripts/install-sdk.sh && ./scripts/install-sdk.sh

# Build the project
pnpm run build

2. Run in local development mode (example):

pnpm run dev

3. Alternatively, run via npx (quick start):

npx -y @klever/mcp-server

4. If you plan to use the public hosted server, follow the Public MCP Server setup and integration steps in the README to connect via Claude, Claude Desktop, Cursor, or VS Code configurations.

Tips and notes:

• Storage mode: memory is best for development and testing; Redis is available for persistent storage and scalable queries. Configure STORAGE_TYPE in .env accordingly.
• When running in public mode, certain write operations are disabled to protect the shared service.
• For production deployments, expose the MCP server behind a reverse proxy with TLS termination and ensure proper handling of mcp-session-id headers and SSE support.
• The knowledge base is large and may be refreshed; in Redis mode, ingesting can be performed once to populate the store.
• Environment variables in .env should reflect your deployment targets (MODE, PORT, STORAGE_TYPE, REDIS_URL, etc.).