LangChain

LangChain added human-in-the-loop to Managed Deep Agents in public beta, letting agents ask users questions through buttons in a Slack thread.

@langchain on X

LangChain added support for NVIDIA's open-weights Nemotron 3.5 Super VL in Deep Agents through ChatNVIDIA, bringing vision, reasoning and tool use.

@langchain on X

LangChain's LangSmith now receives OpenTelemetry traces from Claude Platform, including Managed Agents, the Messages API and the Claude SDKs, in public beta.

@langchain on X

LangChain added support for ScaleDown's constrained decision models, usable with LangChain and Deep Agents and traced through the LangSmith LLM Gateway.

@langchain on X

LangChain launched AI Search in LangSmith, letting users describe what they want to find and get back the matching traces.

@langchain on X

LangChain added support for the OpenAI decisions API, with yes/no, choice and score questions, plus model routing and auto mode middleware.

@langchain_oss on X

LangChain revamped how skills work in Deep Agents to give agents domain knowledge.

@langchain on X

LangChain released Managed Deep Agents v0.9 with a Schedules SDK for agent reminders, per-run model and tool config, and Slack reactions.

@langchain on X

LangChain added free web search, powered by Parallel, to Managed Deep Agents with no separate vendor accounts or API keys needed.

@langchain on X

LangChain's LangSmith Engine v2 now reproduces agent issues, tests and validates fixes automatically, and hands users a ready-made pull request to review.

@langchain on X

LangChain added user-level memory to Managed Deep Agents, so agent memory can be scoped to one user or a whole team.

@langchain on X

LangChain launched Custom Apps in LangSmith, letting users build and publish their own interface on LangSmith data without handling hosting, auth, or permissions.

@langchain on X

LangChain launched LangSmith Fine-Tuning and the smithtune CLI, which turn an agent's own trajectories into a fine-tuned model without a hand-built training pipeline.

@langchain on X

LangChain released MCP-adapters 2.0 for TypeScript, adding support for stateless MCP, tools that ask users for input, and easier auth setup.

@langchain on X

LangChain's Managed Deep Agents 0.8 added HTTP channels, so agents can take requests from any webhook instead of only Slack.

@langchain on X

LangChain added interrupts with typed response schemas to LangGraph, so human replies are validated and apps can show a real response form.

@langchain on X

LangChain added a terminal interface to Managed Deep Agents, started with the mda chat command, for testing agents locally.

@langchain on X

LangChain added Trajectories to LangSmith, a readable chronological view of an agent session for debugging, review, online evals and fine-tuning datasets.

@langchain on X

LangChain released LangSmith Engine v2, which red teams agents before production, validates fixes in the user's environment, and detects inefficient agent work and metric trends.

@langchain on X

LangChain released LangSmith Engine v2, which spots agent issues before users see them and adds red teaming and validated fixes.

@langchain on X

LangChain released Managed Deep Agents 0.8 with user-owned credentials and memory, HTTP channels, native Slack file transfers and built-in web search.

@langchain on X

LangChain added Trajectories to LangSmith, a view of agent behavior that shows an agent session's messages in order for the whole team.

@langchain on X

LangChain launched LangSmith Fine-Tuning and the smithtune CLI in public beta, letting teams train specialized models from their agent traces.

@langchain on X

LangChain made LangSmith Custom Apps generally available, letting users create any interface from their agent data with a prompt.

@langchain on X

LangChain's Managed Deep Agents 0.8 added user memories, with control over where they are stored and no sharing of memories made in private contexts.

@langchain on X

LangChain released OpenWiki 0.6.0, which adds shared workspaces for wikis, faster code wiki generation and support for Antigravity and oh my pi.

@langchain on X

LangChain added support in LangSmith for decision models including Jev and SemIf, giving users visibility into each step.

@langchain on X

LangChain made Jev-as-a-judge available in LangSmith so users can score every production trace and check more criteria per trace.

@langchain on X

LangChain rebuilt trace filtering in LangSmith so users can build precise queries, find exact runs, and see why a result matched.

@langchain on X

LangChain launched Deep Life Sci, an open source agentic assistant for clinical and lab scientists, built on Deep Agents.

@langchain on X

LangChain Managed Deep Agents now support custom middleware for adding behavior around model calls, tool calls, and the agent lifecycle.

@langchain on X

LangChain's OpenWiki v0.5.2 adds a Kiro integration, giving users another coding agent to install it with.

@langchain on X

LangChain added a capstone to its LangSmith Essentials course on LangChain Academy, with more chances to practice debugging and tracing.

@langchain on X

LangChain made every Managed Deep Agent an MCP server, so coding assistants and other agents can call deployed agents through a built-in /mcp endpoint.

@langchain on X

LangChain Managed Deep Agents can now connect to Slack, where mentions, DMs, and thread replies start a run and the agent replies in the conversation.

@langchain on X

LangChain changed how Deep Agents reads files, cutting edit_file errors by 15% and total input token use by 10% in its evals.

@langchain on X

LangChain's Managed Deep Agents now ships with Harbor for running evals in a fresh container each time, tracked in LangSmith.

@langchain on X

LangChain joined the AWS Activate program, giving eligible startups up to $10,000 in LangSmith credits for building and running agents on AWS.

@langchain on X

LangChain's Deep Agents subagents can now fork the supervisor's full conversation instead of starting with a blank context window.

@langchain on X

LangChain added Connections to Managed Deep Agents 0.7, letting agents use shared credentials or each user's own OAuth login, stored in LangSmith.

@langchain on X

LangChain updated its MCP support to cover the new stateless MCP spec, now built into the main langchain package.

@langchain on X

LangChain's OpenWiki now integrates with Cursor, so developers can install it with one command and drive wiki generation from the editor.

@langchain on X

LangChain released OpenWiki 0.5.0, which makes wiki init and update jobs durable and resumable so a failed long run keeps its progress.

@langchain on X

LangChain is a launch partner for Slack's Add to Slack, letting users build an agent in natural language and deploy it to Slack in one click.

@langchain on X

LangChain's Managed Deep Agents can now bake an agent's environment at deploy time from a Dockerfile, so every new thread starts ready.

@langchain on X

LangChain's OpenWiki 0.4.0 adds OKF v0.2 support, so readers can see what sources a page was built from and whether it is verified.

@langchain on X

LangChain's OpenWiki 0.4.0 adds coding agent integrations, so developers can run wiki init and update from Claude, Codex or OpenCode.

@langchain on X

LangChain updated LangSmith Engine with better issue detection, self-hosted support, a cheaper Reduced Analysis mode, Slack and Linear integrations, and automatic closing of stale issues.

@langchain on X

LangChain's Managed Deep Agent deploy now sets up the Slack app automatically, so users need no manifest, OAuth redirects or bot tokens, just one command.

@langchain on X