Plus a bunch of QOL cleanup to make the adapter easier to build with:
— MultiServerMCPClient → MCPAdapter, mcpServers → servers, getTools() → listTools() (old names still work)
— Tool names carry the server name, so two servers can both expose search
— Tool errors from the server reach the model as error ToolMessages
Authenticating against your MCP servers got a lot easier too. Pass an `authProvider` and the adapter handles OAuth for you. It stores tokens, attaches them to requests, and refreshes them when a server sends back a 401.
MCP servers can now pause a tool call to ask the user for more, like a date for a booking or a link to open for payment. It uses LangGraph interrupts, the same human-in-the-loop flow thats used everywhere else in LangChain.
The run gets checkpointed while it waits, so the user can answer days later and the tool call picks up right where it left off.
@𝚕𝚊𝚗𝚐𝚌𝚑𝚊𝚒𝚗/𝚖𝚌𝚙-𝚊𝚍𝚊𝚙𝚝𝚎𝚛𝚜 now supports the new stateless MCP spec, so there's no session to set up or keep alive between tool calls.
Servers that only speak the legacy protocol still work. 𝙼𝙲𝙿𝙰𝚍𝚊𝚙𝚝𝚎𝚛 figures out which version each server supports, or you can pin one with `mode: "modern"` or `mode: "legacy"`.
@𝚕𝚊𝚗𝚐𝚌𝚑𝚊𝚒𝚗/𝚖𝚌𝚙-𝚊𝚍𝚊𝚙𝚝𝚎𝚛𝚜 2.0 is out! 🧵
Using MCP with your TypeScript agents up to MCP servers just got a lot simpler: support for the latest stateless version of MCP, tools that can check in with your users when they need something, much easier auth integrations, and a ton of QoL updates.
Here's what's new 👇
Starting today, @LangChain’s Managed Deep Agents can bake your agent’s environment at deploy time. 🧑💻
No clone. No install. No repeated setup.
Drop in setup.sh or a Dockerfile. Bake once. Every new thread starts ready. 👇
Starting today, a @LangChain's Managed Deep Agent deploy provisions your Slack app for you 🤯
No manifest. No OAuth redirects. No bot tokens to copy around.
One command, and your agent says 👋 in Slack.
An agent is 3️⃣ layers.
👉 Business logic you write.
👉 A harness that runs the loop.
👉 Infrastructure that survives production.
@hwchase17 breaks the whole stack in one video.
youtu.be/TUJmfeGTr1Q
Managed Deep Agents is just getting started 🚀
@hwchase17 just published a set for banger videos that get you up to speed how shipping production agents today looks like.
In Managed Deep Agents skills load in two steps, which is why a lean SKILL.md works.
The prompt keeps the name and description. The body opens when the task matches.
docs.langchain.com/langsmith…
Agent skills fail when their instructions are vague and their context is bloated.
Best Practices for Creating Agent Skills is a practical guide for builders writing skills that agents can discover, load, and execute reliably.
It helps you structure skills around lean SKILL.md instructions, just-in-time references, deterministic scripts, and LLM-assisted validation instead of piling everything into one prompt.
Key features:
• Standard skill structure – separates core instructions, scripts, references, and reusable assets
• Discoverable metadata – explains naming rules and trigger-focused descriptions, including negative triggers
• Progressive disclosure – keeps SKILL.md under 500 lines and loads supporting context only when needed
• Procedural guidance – favors numbered steps, concrete templates, consistent terminology, and third-person imperative
• Validation workflow – covers discovery tests, logic simulation, edge-case testing, and architecture refinement
Free public GitHub repo.
Link in the reply 👇
Mention the agent in Slack. It runs. It replies in the thread.
Managed Deep Agents treats Slack as a channel file: channels/slack.py. Mentions, DMs, and follow-ups start a run as the caller and post the answer back. No bot server to stand up either.
docs.langchain.com/langsmith…
You don't stand up an agent. You upload a folder.
Claude Code is a harness for your laptop. Managed Deep Agents is a harness for production.
Want Slack? Add a file. Want a daily run? Add a file. Want memory? Add a file.
LangSmith runs the harness.
docs.langchain.com/langsmith…
We’re launching Managed Deep Agents today 🎉👀
langchain.com/blog/managed-d…@LangChain
The goal posts for building agents have moved.
Giving an agent tools, deploying it somewhere, and putting a UI in front of it isn’t the hard part anymore.
The next challenge is everything around the agent: identity, memory, credentials, permissions, and securely connecting it to the services your users already use.
This beta release is just the beginning and our foundation what is about to come next 🚀