AgentRPC

Connect AI agents and workflows to any function, deployed anywhere

Freemium

Description

AgentRPC serves as a universal RPC layer designed to unify tools from multiple programming languages, such as TypeScript, Go, and Python (with .Net support planned), into a singular interface for AI agents. It simplifies the integration process by offering automated service discovery and routing, making functions universally discoverable by authenticated clients without needing hardcoded endpoints or custom servers. The platform is compatible with various MCP Servers and OpenAI-Compatible Tool Calling SDKs, including OpenAI, Anthropic, Groq, and LiteLLM.

This middleware facilitates connecting AI agents and workflows to functions deployed in diverse environments, crucially including those within private VPCs, without requiring open ports. AgentRPC manages the communication link; tools execute on the user's own infrastructure while AgentRPC handles registration, polling for calls via its managed platform (api.agentrpc.com), and invocation. This architecture supports long-running functions exceeding standard HTTP timeouts and provides observability features like tracing, metrics, and events, along with automatic failover and retries for enhanced reliability.

Key Features

Multi-language Support: Connect Agents to tools written in TypeScript, Go (.Net coming soon).
Private Network Support: Register and call functions located within private VPCs without needing open ports.
Long-running Functions: Execute tools and APIs that exceed standard HTTP timeout limits using long polling.
Full Observability: Gain complete visibility with tracing, metrics, and events via a hosted platform.
Automatic Failover: Benefit from platform-managed function health tracking, automatic failover, and retries.
Framework Compatibility: Utilize out-of-the-box support for MCP and OpenAI SDK compatible agents.

Use Cases

Integrating AI agents with external tools hosted on different services.
Connecting LLM automations to functions running within private VPCs.
Building distributed AI systems requiring tool calling across network boundaries.
Extending AI agent capabilities without exposing private services publicly.
Simplifying multi-language tool management for AI applications.

Frequently Asked Questions

Why would I want to use AgentRPC?

Use AgentRPC if you're building an AI agent or LLM automation that needs to call external tools, especially if those tools are hosted on different services or in private VPCs. It avoids exposing private services or building custom integrations by letting you register functions via an SDK and get an OpenAI-SDK compatible tool definition.

What advantages does AgentRPC provide over plain old HTTP APIs?

AgentRPC eliminates the need to build custom APIs and service discovery mechanisms. Functions become universally discoverable without hardcoding endpoints. It supports long-running tools beyond HTTP timeouts using async polling and co-locates tool definitions with the function code, generating SDK-compatible definitions automatically. Other benefits include caching and automatic retries.

Where are Tools executed?

Tools are executed on your own infrastructure. AgentRPC acts as middleware, facilitating communication between your agent and the function running on your infrastructure.

How can I connect services that sit behind a firewall or a private network?

When you register a function, the AgentRPC SDK initiates long polling requests to api.agentrpc.com to check for new calls. This outbound connection model means no open ports or firewall configurations are required on your end.

What is AgentRPC not good for?

AgentRPC is not ideal if your agent can call functions directly within the same process. It's designed for agents needing to call external functions in distributed environments. Also, use cases with very strict low-latency requirements might not tolerate the additional network hop through AgentRPC.