LangSmith Launches Fleet for Enterprise Agents

🔑 Enhanced Key Takeaways

• •Fleet enables non-technical teams to create no-code agents for tasks like daily briefings, competitor tracking, and project updates by simply describing needs, with the platform building and iteratively improving agents based on feedback.
• •Fleet incorporates human-in-the-loop approvals, background agents, and multi-agent coordination on a durable runtime ensuring exactly-once execution for handling real-world enterprise interactions.
• •As part of LangSmith's enterprise platform, Fleet integrates with NVIDIA technologies including Nemotron models, NeMo Agent Toolkit, NIM microservices, and OpenShell for secure, sandboxed agent runtime with policy-based guardrails.
• •LangSmith, powering Fleet, has processed over 15 billion traces and 100 trillion tokens, offering observability features like distributed tracing, Insights Agent for pattern detection, and Polly for natural-language debugging.

🛠️ Technical Deep Dive

• •Fleet supports deployment with versioning, rollbacks, and native protocols like A2A, MCP, and Agent Protocol for standardized enterprise-wide agent management.
• •Integrates LangGraph for stateful, multi-actor cyclic workflows coordinating multiple chains and agents.
• •LangSmith Evaluation enables offline evals (LLM-as-judge, human review, pairwise comparison, CI/CD via pytest/GitHub) and online multi-turn evals scoring conversation trajectories.
• •NVIDIA integration includes Nemotron models, NeMo Agent Toolkit for profiling/optimization, NIM microservices, Dynamo for improvement, and OpenShell for secure sandboxed runtime.

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline