Hi Product Hunt, I’m excited to share Tern, an open-source interoperability framework for Coding Agents and Language Models. The AI coding ecosystem is moving extremely fast. New agents appear regularly, model capabilities change quickly, local inference is becoming more practical, and deployment requirements vary widely between teams. That creates a problem: once your workflow is deeply tied to one agent or one model provider, switching becomes expensive. Tern is our attempt to solve that. Instead of building directly around a single coding agent or LLM provider, Tern provides a common layer for running agents and routing model calls. It currently supports Claude Code and Codex as coding agents, and OpenAI, Anthropic, Google, and Ollama as LLM backends. The project includes: A server for running coding agent sessions A CLI client, ternctl A Go client library Claude Code and Codex adapters LLM backends for OpenAI, Anthropic, Google, and Ollama Secure key vault support Early protocol work around CAWA and LLMGP The long-term vision is context portability: developers should be able to move between agents, models, and deployment environments without throwing away their workflow or starting from scratch. Tern is still early, and we are looking for feedback from developers, AI tool builders, and teams experimenting with coding agents in real projects. I’d love to hear: Which coding agents or LLM backends should we support next? What would make agent/model switching actually useful in your workflow? How should context portability work in practice? Thanks for checking it out.

Tern

Open-source interoperability layer for Coding Agents & LLMs

Open-source interoperability layer for Coding Agents & LLMs