← Back
LangChain
LangChain releases LangSmith CLI and skills to boost AI agent performance by 75 percentage points
· releasefeatureapisdkopen-source · blog.langchain.com ↗

LangSmith CLI: Agent-Native Infrastructure

LangChain released a new LangSmith CLI designed from the ground up for agent-native workflows. The CLI provides the building blocks needed to programmatically interact with the LangSmith platform—including fetching traces, curating datasets, and running experiments. By enabling terminal-first access to LangSmith capabilities, the CLI positions itself as critical infrastructure for agent-driven development loops where agents can improve other agents through automated iteration.

Skills: Dynamic, Domain-Specific Guidance

The release also introduces Skills, a mechanism for enhancing AI agent performance in specialized domains. Unlike traditional approaches that burden agents with all available tools at once (which historically degrades performance), skills are dynamically loaded through progressive disclosure—agents only retrieve instructions when they're relevant to the current task. Skills are portable and shareable, consisting of markdown files and scripts that can be retrieved on demand.

LangSmith Skills: Tracing, Datasets, and Evaluation

LangChain is shipping three foundational skills covering the core areas of LangSmith AI engineering:

  • Trace: Add tracing to existing code and query execution traces
  • Dataset: Build datasets of examples for testing and validation
  • Evaluator: Assess agent performance across those datasets

These skills enable a virtuous cycle: agents can add tracing to their own code, generate traces during execution, convert those traces into test datasets, create evaluators, and iterate on their architecture based on evaluation results.

Benchmark Results and Installation

On LangSmith-focused tasks, Claude Code's pass rate improved dramatically—from 17% to 92% when equipped with the skills. LangChain plans to open-source the testing benchmark used for these evaluations.

Users can install the LangSmith CLI via:

curl -sSL https://raw.githubusercontent.com/langchain-ai/langsmith-cli/main/scripts/install.sh | sh

Skills can be installed using npx skills:

npx skills add langchain-ai/langsmith-skills --skill '*' --yes

The release is part of a broader effort to enable AI agents to build, test, and improve other agents—a paradigm shift LangChain expects to define agent development going forward.