The AI SDK Harness API now supports @OpenCode and @LangChain Deep Agents through a single unified interface.
Use 𝙷𝚊𝚛𝚗𝚎𝚜𝚜𝙰𝚐𝚎𝚗𝚝 to run any supported runtime without changing your app code. https://vercel.com/changelog/deepagents-and-opencode-harness-adapters
Positive users expressed excitement about Vercel AI SDK adding support for LangChain and OpenCode agents, while negative users called the adapter awkward for cloud use cases and the framing weak for making claims hard to verify.
No Digg Deeper questions have been answered for this story yet.
@vercel_dev @opencode @LangChain just in time for us to completely change things with our v2 api!
💪deep agents harness
The AI SDK Harness API now supports @OpenCode and @LangChain Deep Agents through a single unified interface.
Use 𝙷𝚊𝚛𝚗𝚎𝚜𝚜𝙰𝚐𝚎𝚗𝚝 to run any supported runtime without changing your app code. https://vercel.com/changelog/deepagents-and-opencode-harness-adapters
@vercel_dev @opencode @LangChain just support acp and we can take care of it from there
@thdxr @vercel_dev @opencode @LangChain No worries, my agent doesn't get mad at you for giving it more work
@vercel_dev @opencode @LangChain @thdxr mentioned
@yousefrol @vercel_dev @opencode @LangChain We have an adapter but - it's the most awkward - it's not a good fit for the cloud deployment use case for now. It seems designed for the "editor on my laptop" use case - I this the ACP team is aware and looking at updating the protocol
@vercel_dev @opencode @LangChain hype
@vercel_dev @opencode @LangChain third unified agent layer i've tried this year
@cramforce @vercel_dev @opencode @LangChain super cool 🤍
x.The unified interface is nice, but the real test is how it handles state and memory serialization when you actually try to swap them. The abstractions usually leak once the agent gets complex.The unified interface is nice, but the real test is how it handles state and memory serialization when you actually try to swap them. The abstractions usually leak once the agent gets comple
@yousefrol @vercel_dev @opencode @LangChain deep agents has you covered https://docs.langchain.com/oss/python/deepagents/acp
@hwchase17 Weak framing — "harness" without specifying what’s being harnessed (compute, data, or tooling?) makes the thesis hard to verify. Need a clearer bottleneck or metric.
the quiet story in agent infra right now: everyone's racing to make the runtime not matter.
swap opencode for langchain, never touch your app code. the agent layer is becoming a commodity you plug in and out.
when the runtime stops being the moat, whatever the agent actually does in the real world becomes the whole game.