LangSmith adds Engine v2, Managed Deep Agents 0.8 and fine-tuning

Engine v2 finds issues before production and tests fixes

LangSmith Engine is an in-platform agent that automates steps in the agent development lifecycle, from spotting issues to proposing and testing fixes. LangChain says that since launching in May, Engine has helped engineers analyze more than 60M traces and diagnose tens of thousands of issues.

Engine v2 adds Red Teaming. Using production traces and repos, Engine can generate hypotheses about issues that may not have appeared in production yet, test them, and surface confirmed failures for review.

Issue detection is also broader. Beyond errors and unmet user requests, Engine can now identify performance trends across error rate, latency and cost, as well as inefficient agent work such as repetitive tool calls or unnecessarily long trajectories.

For agents running on LangSmith Deployment, Engine can validate proposed fixes before human review. It runs the offending inputs against the agent to confirm the issue, then tests candidate fixes against a broader eval set until one resolves it. The user can then open a PR with a click. The next release of self-hosted LangSmith will support BYOK for Engine, according to LangChain.

Managed Deep Agents v0.8 adds user memory, channels and web search

Managed Deep Agents combines the Deep Agents harness with managed infrastructure. Version 0.8 addresses agent memory, authentication, channels and tool management.

• User-level memory: agents can store context scoped to each authenticated user, separate from shared agent-level memory. The runtime does not copy content between layers, and access policies can be set at each level.
• Credentials: both agent-owned and user-owned credentials are supported for external connections such as GitHub and Notion.
• Channels: Slack support now includes file transfer, and a new HTTP channel lets teams connect agents to any service that can send a JSON webhook.
• Web search: a prebuilt tool powered by Parallel, which removes the need for a separate vendor account, API key or custom tool.

Trajectories and Fine-Tuning turn sessions into training data

Long-running agent sessions can include many turns, tool calls, retries and subagent handoffs, which makes full traces long and nested. LangSmith Trajectories is a conversational view that aggregates messages from humans, AI and tools across the main agent and any subagents, in chronological order. LangChain says this speeds up debugging and gives subject matter experts a readable view they can score, flag and annotate without parsing execution metadata. LangSmith online evaluators can also score Trajectories.

LangSmith Fine-Tuning builds on this. It supports supervised fine-tuning (SFT) of open models on high quality examples of a task. The CLI, smithtune, builds and prepares datasets from trajectories, trains the model with Baseten or Fireworks, evaluates results with LangSmith, and serves tuned models and connects them to applications. LangChain argues that a model trained on a team's own data can often perform specified tasks as well or better than a general purpose frontier model, at a much lower cost and latency. The company says it benchmarked SmithTune internally and published a full breakdown, but the announcement does not give figures.

Custom Apps and what it means for teams

Custom Apps lets teams build, publish and run custom interfaces on top of LangSmith data. Teams can build against LangSmith APIs using templates, in-app chat or coding agent guidance, then publish the app into their workspace as a shared interface. LangChain points to annotation workflows, experiment comparisons across versions or slices, and trace review as use cases.

Taken together, the releases are framed by LangChain as supporting its mission to let every company own its intelligence, with customized, domain-specific agents tailored to specific problems. For teams already on LangSmith, the changes cover detecting and fixing problems, running agents with per-user context, and reusing session data to tune smaller open models. The announcement does not state pricing or availability beyond the features described.