solana

The Solana MCP Server exposes a comprehensive MCP JSON-RPC interface for querying Solana blockchain data via natural language conversations. It supports two operation modes: stdio (default) for integration with Claude Desktop and other MCP clients, and a web service mode that serves an HTTP JSON-RPC API. In stdio mode, you run the server with solana-mcp-server stdio, which allows conversational agents to invoke Solana RPC methods behind the scenes. In web mode, you run solana-mcp-server web (optionally with a custom port) to expose endpoints such as POST /api/mcp for the MCP JSON-RPC API, GET /health for health checks, and GET /metrics for Prometheus metrics. The server implements a wide range of Solana RPC methods (account, block, and other networks-related calls) and includes an internal RPC caching layer to improve latency.

When configuring Claude Desktop or another MCP client, provide the mcpServers configuration so that the client knows how to launch the server and connect to the intended RPC endpoint. The server can be pointed at your Solana RPC endpoint via the SOLANA_RPC_URL environment variable. This makes it straightforward to switch between mainnet, testnet, or custom RPC servers as needed. Tools exposed by the MCP include the standard MCP JSON-RPC methods such as getAccountInfo, getBalance, getBlock, getBlockTime, and many more, all backed by Solana's RPC surface. The built-in caching and metrics help you monitor performance and adjust TTLs for frequently used methods.

To use the web API, send HTTP POST requests to the /api/mcp endpoint with MCP JSON-RPC payloads. You can also query /health for a quick capability and health summary and /metrics to observe RPC activity and cache statistics. When testing locally, you can start in stdio mode during development and switch to web mode for integration with web applications or automation workflows.

Prerequisites:

• A Unix-like environment (Linux/macOS) or Windows with a compatible shell
• curl or wget for downloads
• Rust toolchain if building from source (cargo)
• Git for cloning the repository (optional if using pre-built binaries)

Option 1: Quick one-liner installation (Solana CLI approach used for Claude Desktop)

1. Run the installer script to fetch pre-built binaries and configure Claude Desktop:

curl -fsSL https://raw.githubusercontent.com/opensvm/solana-mcp-server/main/scripts/install.sh | bash

This script will download binaries (or build from source if needed), configure Claude Desktop, set up environment variables (including SOLANA_RPC_URL), and back up existing configurations.

Option 2: Alternative install methods

• Using wget:

wget -qO- https://raw.githubusercontent.com/opensvm/solana-mcp-server/main/scripts/install.sh | bash

• Manual download and run:

curl -fsSL https://raw.githubusercontent.com/opensvm/solana-mcp-server/main/scripts/install.sh -o install.sh
chmod +x install.sh
./install.sh

Option 3: Manual build from source (advanced)

1. Clone the repository and build from source:

git clone https://github.com/opensvm/solana-mcp-server.git
cd solana-mcp-server
cargo build --release

2. After building, configure Claude Desktop to point to the built binary, e.g. by setting the command path in your config.json to the path of target/release/solana-mcp-server and ensuring SOLANA_RPC_URL is set.

Notes: After installation, restart Claude Desktop and start querying Solana data directly. You can switch between stdio and web modes as needed. If you prefer a containerized deployment, you may adapt the configuration to run solana-mcp-server inside Docker or Kubernetes once you have a compatible image available.

Environment variables and configuration tips:

• SOLANA_RPC_URL: The RPC endpoint used to fetch Solana data. Replace with mainnet-beta, testnet, or a private RPC as needed.
• For caching, consider TTL overrides per method to balance freshness and latency. The server supports per-method TTL settings (refer to the caching documentation in docs/caching.md).
• In web mode, you can customize the HTTP port with --port (e.g., solana-mcp-server web --port 8080).
• Ensure proper networking access if you run behind a firewall or in a Kubernetes cluster.
• If you encounter RPC timeouts or rate limits, tune the caching layer to cache frequently requested data and adjust backoff settings.
• The server exposes Prometheus metrics at /metrics; you can scrape and import them into your monitoring stack.
• For Claude Desktop integration, you typically only need to provide the solana mcpServer configuration, including the command path and environment variables, so that Claude can launch and connect to the server automatically.