Loading tool details…
AgentRPC: Open-source RPC framework (Apache 2.0) that lets AI agents call functions across network boundaries without opening ports. Supports TypeScript, Go, and Python SDKs with built-in MCP server compatibility.
Open-source RPC framework that lets AI agents call functions behind firewalls and across private networks without opening ports — supports TypeScript, Go, and Python with built-in MCP server compatibility.
AgentRPC is a free, open-source AI agent RPC framework (Apache 2.0) with optional managed hosting (quote-based pricing) that lets AI agents call functions across private network boundaries without opening inbound ports. It solves a specific infrastructure problem: when your AI agent runs in one environment but needs to invoke functions inside a private VPC, Kubernetes cluster, or firewalled network, AgentRPC's outbound-only SDK connections bypass the need for VPNs, open ports, or ingress configuration entirely.
The framework ships native SDKs for 3 languages — TypeScript, Go, and Python — with a 4th (.NET) in active development as of early 2026. Each SDK uses long-polling to keep connections alive beyond the standard 30–60 second HTTP timeout, which is critical for agent tasks involving database queries, report generation, or multi-step data processing that run for several minutes. The TypeScript SDK includes a built-in MCP server that exposes registered functions to Claude Desktop, Cursor, and any MCP-compatible host without additional configuration, and all SDKs generate OpenAI-compatible tool definitions that work with Anthropic, LiteLLM, and OpenRouter.
The managed hosting option at api.agentrpc.com adds observability features including tracing, metrics, structured event logging, and health monitoring for registered functions — pricing is quote-based and scales with function call volume and team size. Self-hosting is fully supported with zero software licensing costs under Apache 2.0.
AgentRPC launched publicly in Q4 2025 and has accumulated over 870 AI tool integrations across its directory. It is purpose-built for cross-network agent function calling rather than general workflow orchestration, making it narrower in scope than platforms like Temporal or Inngest but significantly simpler to deploy for its target use case. The GitHub repository at github.com/agentrpc/agentrpc contains all SDK source code, documentation, and example implementations for each supported language.
AgentRPC solves a specific problem: letting AI agents call functions across private networks without VPNs or open ports. Valuable for teams with private VPC infrastructure who need AI agent access to internal functions. The SDK approach with 3 language SDKs (TypeScript, Go, Python) and built-in MCP server compatibility makes integration straightforward for developer teams. The managed hosting adds observability but introduces latency overhead. Best suited for teams with genuine cross-network function calling needs rather than as a general-purpose RPC replacement.
Register functions from private VPCs, Kubernetes clusters, or firewalled environments without opening inbound ports. The SDK establishes an outbound connection to the RPC server, making it compatible with restrictive network policies. This eliminates the need for VPNs, port forwarding, or ingress configuration that traditional RPC frameworks require.
Standard HTTP requests time out after 30-60 seconds, which breaks agent tasks involving database queries, report generation, or multi-step data processing. AgentRPC uses long-polling to keep connections alive for minutes, letting agents wait for results from functions that take extended periods to complete.
Native SDKs ship for TypeScript, Go, and Python today, with .NET in active development. Each SDK exposes the same registration API and health-check patterns, so polyglot teams get consistent cross-language function exposure through a single RPC layer without writing glue code between stacks.
The TypeScript SDK includes an embedded MCP server that exposes registered functions to any MCP-compatible host, including Claude Desktop and Cursor, without additional configuration. This eliminates the need to write a separate MCP server adapter for internal tools.
The hosted platform provides tracing, metrics, and structured event logging for every function invocation. Health tracking continuously monitors registered functions, surfacing issues like disconnects, error rates, and latency degradation in real time.
Free
Quote-based (contact for pricing)
Ready to get started with AgentRPC?
View Pricing Options →npm install @agentrpc/sdk. For Python: pip install agentrpc. For Go: go get github.com/agentrpc/agentrpc.
We believe in transparent reviews. Here's what AgentRPC doesn't handle well:
.NET SDK in active development as of early 2026, expanding language coverage beyond the current TypeScript, Go, and Python SDKs. No public release timeline has been announced.
Workflow Orchestration
Durable execution platform for AI agents and workflows — used by Netflix, Snap, and Stripe for reliability-critical systems.
AI Infrastructure
Serverless cloud for running Python code, AI models, and GPU workloads at scale with per-second billing.
AI Workflow Orchestration
Durable execution engine that automates workflows and AI agent runs — event-driven functions with retries, flow control, and step-level observability with no infrastructure to manage.
AI Automation
Source-available workflow automation platform with native AI agent and LangChain nodes, popular for self-hosted AI workflows.
Pricing and features change often. If anything on this AgentRPC page is wrong or missing, tell us and include a link to the source. We review every correction.
No standalone guide download is available here.
Use our existing overview and feature information to evaluate AgentRPC.