◆ FOR AGENTS

Built for agents, not just humans.

On-chain agents don't read hero sections - they read tool definitions and structured files. rpc edge is machine-native: an open MCP server plus skills.md, llms.txt, and machine-readable pricing, so your agent can discover, evaluate, and use it programmatically.

◢ Get the MCP serverWhy agents need speed →

◆ THE MCP SERVER

Solana infra, as tools.

solana-infra-mcp is an open, provider-agnostic MCP server (rpc edge as the flagship backend). Add it to Claude Code or any MCP client:

git clone https://github.com/rpc-edge/solana-infra-mcp
cd solana-infra-mcp && pnpm install && pnpm build
claude mcp add solana-infra -- node "$(pwd)/dist/index.js"

// tools: rpc_health · priority_fee_estimate · epoch_info · next_leaders · latency_compare · submit_transaction

◆ MACHINE SURFACES

Readable by machines.

/skills.md

Action-oriented integration guide: get a key, authenticate, endpoints, quickstart snippets in TS/Python/Rust/curl.

/llms.txt

The llms.txt index of the site - products, pricing, guides, reference - for LLMs and agents.

/llms-full.txt

Full content in one file: endpoints, product descriptions, glossary, comparisons, FAQ.

/pricing.md

Machine-readable pricing so an agent can evaluate cost without scraping a page.

◆ AGENT FAQ

Straight answers.

Do you have an MCP server for Solana?
Yes - solana-infra-mcp is an open (MIT) MCP server that exposes Solana infrastructure as tools: RPC health, priority-fee estimation, epoch/slot state, the leader schedule, latency comparison, and a keyless signed-transaction relay with landing confirmation. It is provider-agnostic with rpc edge as the flagship backend, so an agent can query Solana infra directly.
Why does an AI agent need agent-ready infrastructure?
Agents evaluate and integrate a provider through tool definitions and structured files, not a marketing page. A provider whose pricing and capabilities are only human-readable gets filtered out of agent-mediated selection. rpc edge ships /skills.md, /llms.txt, /llms-full.txt, /pricing.md, and an MCP server so an agent can discover, price, and call it programmatically.
How does an agent authenticate to rpc edge?
The same ways any client does: an API key as a query parameter or header on HTTP/WebSocket, and as metadata on the Yellowstone gRPC stream. The MCP server reads the endpoint from its environment (SOLANA_RPC_URL), so the key stays out of prompts and logs. See /skills.md for the exact auth forms.
◢ Talk to the teamRead skills.md →

// provider-agnostic, MIT-licensed · rpc edge as the flagship backend · last updated 2026-07-13