Grafana Labs has open sourced the Grafana AI SDK for Go, a shared foundation built internally to stop teams from reinventing provider clients, streaming abstractions, and error handling for every new LLM experiment. It provides common interfaces for calling models, streaming responses, executing tools, generating structured output, and running multi-step agents, and it speaks the Vercel AI SDK's UI message stream protocol so a Go backend can stream directly to React hooks like useChat, useCompletion, and useObject. It currently supports Anthropic, Amazon Bedrock, OpenAI's Responses API, and OpenAI-compatible APIs, with provider integrations split into separate Go modules. The SDK is explicitly modeled on Vercel's AI SDK, tracks a public compatibility baseline against pinned upstream versions, and ships with Agent Observability middleware, structured logging, and Prometheus metrics support. Grafana frames it as a young, evolving project and invites the Go and open source community to contribute providers, middleware, and feedback.
Table of contents
What teams can build with it todayFrom fast experiments to common patternsInspired by Vercel AI SDK, implemented for GoOne interface across providersQuestions this post answers
What is the Grafana AI SDK for Go and what providers does it support?
It is an open source Go SDK, built by Grafana Labs, that provides shared interfaces for calling LLMs, streaming responses, executing tools, generating structured output, and running multi-step agents. It currently supports Anthropic, Amazon Bedrock, OpenAI's Responses API, and OpenAI-compatible APIs, with each provider shipped as a separate Go module so applications only install what they use. Go developers evaluating AI SDK options for backend LLM work can track releases like this on daily.dev.
How does a Go backend stream LLM responses to a React frontend using useChat?
The Grafana AI SDK implements the UI message stream protocol used by @ai-sdk/react, so a Go handler can write server-sent events directly to useChat, useCompletion, or useObject without a protocol adapter. Tools, reasoning, sources, files, and application data move through typed message parts, letting Go backend teams and React frontend teams share the same AI SDK conventions. Teams wiring Go services to AI SDK React hooks can follow protocol changes like this on daily.dev.
Is the Grafana AI SDK for Go a full port of the Vercel AI SDK?
No, it is not a line-by-line port or a claim of complete feature-for-feature parity. It follows the Vercel AI SDK's model interface, orchestration layer, and frontend wire protocol as a reference for behavior and naming, but some upstream features are not yet implemented in Go, and Grafana maintains a public compatibility baseline and a parity coverage map documenting the known gaps and deviations. Developers comparing AI SDK implementations across languages can weigh trade-offs like this on daily.dev.