AgentRPC

AgentRPC is a universal RPC layer designed for connecting AI agents to functions across network boundaries and different programming languages. It's particularly useful for services deployed in private VPCs, Kubernetes clusters, or multi-cloud environments.

AgentRPC works by wrapping functions in a universal RPC interface and connecting them to a hosted RPC server accessible through the Model Context Protocol (MCP) and OpenAI-compatible tool definitions.

Key features include multi-language support (TypeScript, Go, Python, and .NET), private network support without open ports, long-running function capabilities via long polling SDKs, comprehensive observability with tracing and metrics, automatic failover, and framework compatibility with MCP and OpenAI SDKs.

The AgentRPC platform manages function registration, monitors health, and provides OpenAPI SDK compatible tool definitions and a hosted MCP server. The TypeScript SDK includes an MCP server that can be integrated with Claude Desktop and Cursor using configuration files and API secrets. The project is licensed under Apache 2.0.