ELSEIF
Your brief EB
332 stories from 97 feeds 278 clusters Refreshed 15 minutes ago next pull 15:21

PLATFORMS Signal 284

AI SDK harness layer adds ACP-compatible harness support via meta adapter package

Vercel’s AI SDK now allows integration of any Agent Client Protocol-compatible harness through a new meta adapter package

WHY IT MATTERS

Engineers building AI-powered tools can now reuse existing ACP-compatible harnesses without writing custom adapters for each runtime. This reduces integration overhead but may trade off some runtime-specific optimizations. The change expands compatibility while preserving the option for tighter, direct integrations where needed.

Written by elseif from the cluster below · every claim links back to a source

The three things worth knowing

01

The new @ai-sdk/harness-acp package enables ACP-compatible harnesses to work with the AI SDK harness layer

02

Previously, each harness required a dedicated adapter for a specific runtime like Claude Code or Codex

03

Direct adapters remain preferred for runtimes where ACP support limits functionality or performance

THE READ

What the cluster adds up to.

ORIGINAL ANALYSIS

Vercel’s AI SDK harness layer has introduced a meta adapter that allows any Agent Client Protocol-compatible harness to integrate with the SDK. The @ai-sdk/harness-acp package abstracts the protocol itself, rather than wrapping a single runtime. This means engineers can now implement a harness by passing an ACP-compatible package to `createACP` and configuring basic mappings, such as permission modes and authentication methods. The change simplifies the process of adding support for new or niche harnesses that ship with ACP compatibility but lack a dedicated AI SDK adapter.

The shift to a protocol-based approach reduces the need for custom adapters for each runtime. Previously, the AI SDK required a separate adapter for every harness, such as `@ai-sdk/harness-claude-code` or `@ai-sdk/harness-codex`. While this ensured tight integration, it also created maintenance overhead for engineers who needed to support multiple runtimes. The new meta adapter allows a single implementation to work with any ACP-compatible harness, provided the harness ships an ACP package. This is particularly useful for less common or experimental runtimes that may not justify a dedicated adapter.

Despite the added flexibility, the AI SDK team cautions that ACP-based integrations may not always be the best choice. For runtimes like Claude Code and Codex, direct adapters are still recommended because they can expose runtime-specific behaviors more effectively. ACP, as an abstraction, may limit or alter how certain internal features are surfaced, leading to suboptimal performance or functionality. Engineers will need to weigh the trade-offs between the convenience of ACP compatibility and the potential benefits of a tighter, runtime-specific integration.

The implementation process for the new meta adapter is straightforward. Engineers define a harness by specifying its package source, executable, environment variable forwarding, permission mappings, and authentication method. Once configured, the harness can be passed to `HarnessAgent` like any other adapter. This consistency in usage means existing codebases can adopt ACP-compatible harnesses with minimal changes. However, the documentation notes that not all harnesses support ACP, and some may require direct adapters to unlock their full potential.

Written by elseif from the cluster below · checked for specifics the sources never contained

THE CLUSTER

Same story, 1 feed.

ORDERED BY FIRST SEEN
Vercel Use ACP-compatible harnesses with the AI SDK harness layer Open ↗