LangChain launched Engine, an agent that autonomously investigates traces and drafts PRs to improve other agents.
“We're working towards a future where agents improve themselves.”
langchain
41 tracked signals on langchain.
LangChain launches LangSmith fine-tuning in public beta with SmithTune, a CLI to post-train models from agent traces.
“today we're launching LangSmith fine-tuning in public beta with SmithTune, a CLI to allow you to post-train models from your LangSmith traces in one workflow”
Manually reviewing 100-1,000 real examples beats any automated benchmark for evaluating AI models.
“there's just no better eval than looking at 100 examples or 1,000 examples”
LangChain CEO argues companies must own their agent harness to truly own their AI intelligence
“The main job of a harness is to bring context to the model at the right point in time.”
LangSmith Sandboxes gives AI agents isolated, scalable compute environments spinning up in under a second
“The question isn't whether agents need computers. It's how you give them one, safely.”
LangChain launched dynamic subagents in Deep Agents, letting agents spawn and coordinate parallel subagents by writing code.
“So the orchestration effectively moves out of the agent's head and into code.”
LangSmith Engine and Context Hub turn agent traces into durable, versioned long-term memory for continual learning.
“Most agents don't learn, they just leave traces.”
Lyft scaled to seven+ production AI agents at 35% resolution by building an offline-eval quality gate before shipping.
“You don't want to use your users as test data.”
LangChain launched LangSmith Sandboxes to safely run untrusted agent code with sub-second spin-up.
“agents are writing real code today”
Microsoft Foundry separates agent identity, conversation state, and executable workspace into distinct controls
“Agent memories becoming ordinary cloud data architecture.”
LangChain's Managed Deep Agents enables Slack-to-GitHub PR agents in minimal code
“Having an agent called from Slack is a pretty substantial request, but with Managed Deep Agents channels, this is done with just a couple lines of code.”
LangChain adds cron-based schedules to Managed Deep Agents for recurring, tool-invoking messages delivered to Slack.
“Schedules can be configured and added to your agent to allow it to send recurring or regular messages that invoke its tools or leverage context that it has available.”
LangChain has evolved from an open source framework into a full agent development lifecycle platform anchored by LangSmith.
“Turns out building the agent is really fun. Super fun. And kind of easy now, but actually keeping it from going completely sideways in production is the hard part.”
TypeSafe's Jev is a new 'System 1' model class making fast type-safe decisions rather than generating text.
“Jev does not generate text. It is not a generative AI in that sense... Jev is about making fast decisions from specifically well-scoped questions with type safe schemas for each of them.”
LangChain's Managed Deep Agents let you build a Slack-to-GitHub-PR coding agent in a few lines of code.
“It's an agent that can be called through Slack to write GitHub PRs. And best of all, it only took a few lines of code to write and to deploy.”
LangChain frames agent governance as a new engineering discipline requiring cost, reliability, and compliance controls.
“Production agents introduce a different risk profile than traditional LLM apps. They have greater autonomy, meaning they have a greater need for visibility and control.”
Toyota deploys deep agent systems for R&D research across paint, manufacturing, and supply chain.
“Deep Agents is one beautiful concept where deep agents is like one single command. That's a creative agent.”
LangChain Engine shifted from auto-PR generation to an inbox model for agent bug triage
“We often find that things that are real issues that we can say are objectively problems with the agent, they're just unimportant to a human.”
Use MCP servers for known, enterprise-auth, human-in-the-loop tools; use CLIs for model-familiar, light-surface tasks to save tokens.
“the MCP server for GitHub is about 45,000 tokens because it's giving every single API name, description, and schema to the model. But with the CLI, running the help command where it can see all of the tools, there's only about 1,500 tokens”
Monday.com shifted its mission from managing work to doing the work, using Deep Agents to power Sidekick and an agent suite.
“instead of like just helping company to manage their work with the boards, we want actually to doing the work”
Sandboxes are isolated virtual computers that let agents run arbitrary code safely and scale via parallelization.
“Sandboxes are isolated virtual computers which your agents can use to write code, run commands, or automate browser tasks.”
Benchling makes everyone on the team review AI traces through structured rotations, user feedback, and feature ownership.
“we have a weekly "fire chief" rotation who's addressing issues”
LangChain launches LangSmith Fleet, letting anyone build and manage no-code agent fleets via chat.
“With Fleet, we skip the "if this, then that" canvas entirely and go straight to a fully agentic system that combines significant power with the simplest creation experience: chat.”
Etsy built a production gifting agent on LangChain with a thin ReAct harness that drove high purchase rates.
“What we found was that it returns high-quality search results with a relatively thin harness”
Cisco built agentic systems to manage its $60B+ recurring-revenue customer-experience business and shares lessons learned.
“their mission in life is to maximize the value of the investment that someone did in Cisco”
Foundry lets agents built with open-source frameworks deploy to production without rewriting, using OpenAI-compatible APIs and MCP.
“Changing the model target is configuration not a rewrite.”
Supply-side agents capture org context and decisions without user access by living where work happens.
“you don't really need to have access to the user. What you need access to is the system of context as well as decisions are being made throughout the organization.”
Continual learning lets AI agents update their own prompts, skills, and tools over time from feedback.
“continual learning is when you give the agent the ability to update itself over time, say it's prompts, subagents or skills”
NVIDIA proposes adding specialized deep research skills to agent harnesses like Claude Code
LangChain demos 'Jev,' a fast, cheap System 1 classification model for agent routing, guardrails, and evals.
“System 1 models are a class of AI models built to make fast structured decisions that software can use directly.”
LangChain launches AI-powered coding agent tutors for LangChain Academy courses
LangChain releases Managed Deep Agents with a CLI scaffolding tool for structured agent projects
An agent harness is the surrounding engineering scaffolding built around the core model-and-tool-calling loop.
“Harness is sort of the surrounding support that we add onto that core model and tool-calling loop.”
LangChain shows how to convert an existing LangGraph agent into a voice agent using the Pipecat framework.
“Pipecat is going to handle all of the glue required to take the input audio, convert it to text, run it through our LangGraph LLM layer, then convert that back to speech in order to send it back to the end user.”
AWS shows how to build context-isolated research agents using LangChain Deep Agents and Bedrock AgentCore subagents.
“A better approach is to delegate deep work to isolated subagents that return only concise results.”
LangChain adds support to deploy Google ADK agents directly on LangSmith Deployment.
“You can now deploy a Google ADK agent straight to Langsmith Deployments.”
Odessia uses LangSmith Engine to debug agent traces and democratize agent development across non-expert team members.
“There's this democratization where a lot more people on the team can now contribute.”
A vendor combines AI agent swarms with a high-scale data platform to prioritize enterprise security vulnerabilities.
“agents are actually coloring the graph over time to create more and more interesting lower high confidence edges”
LangSmith is LangChain's framework-agnostic platform for tracing, testing, deploying, and monitoring LLM agents.
“Agents are a black box, making them difficult to observe and debug.”
LangChain's Managed Deep Agents lets you edit agent instructions without redeploying code
LangChain introduces LangSmith Custom Apps for building custom interfaces around agent data