tRPC Agent Go

tRPC Agent Go is an open-source framework for developers building agent services with local automation in Go. It can execute skill scripts in an isolated workspace and includes deployable HTTP servers for agent interfaces. The documented OpenAI and DeepSeek workflows require a provider API key, so running the service yourself doesn't make those model calls local.

The framework suits document research, database assistants and automation that needs several steps or cooperating agents. Graph workflows support conditional branches, while other agent types run tasks sequentially, concurrently or in repeated planning and execution loops. Streaming responses and cancellation support interactive applications.

Reusable SKILL.md procedures let agents run defined tasks through a local executor. Completed sessions can undergo review and quality checks before becoming managed skills for later work. Evaluation supports repeatable test sets, configurable metrics and local file-backed results, though the agent being evaluated can still call a remote model.

Document retrieval workflows load sources, create embeddings and search a vector store. Session and memory services support in-memory storage and Redis. Agents can call Go functions or MCP tools, and A2A connects them to agents in other systems.

The Apache 2.0 framework includes an AG-UI streaming server with CopilotKit and TDesign Chat client examples, plus a minimal Telegram gateway. OpenTelemetry tracing covers model calls, tools and the runner, with trace export for inspecting execution.