←
nathan drezner955 posts
nathan drezner avatar

nathan drezner

@ndrezn

building Managed Deep Agents @langchain. I post demos, code, and what breaks. prev @plotlygraphs nathan (at) drezner (dot) xyz

◷ Joined May 2013▧ 192 Photos and videos
nathan drezner avatar
nathan drezner@ndrezn·Oct 28

nearly every proactive agent you've seen is a cron job with a prompt memory that "dreams" overnight, agents that ping you first. both are scheduled runs on a 50+ year old OS idea how to build one with Managed Deep Agents ⬇️

nathan drezner avatar
nathan drezner@ndrezn·20h

knowing how to build proactive agents means defining how those agents can act on their own. proactive agents can handle - - real-world events, like a new email or a newly opened pull request. - schedules set by the developer, like a nightly job that updates the agent's memory or a weekly sweep for stale pull requests. - schedules set by the agent, like a reminder to follow up on a thread tomorrow, or a daily brief a user asked for. these triggers mean the agent can perform actions far beyond what's possible with a simple chat interface (and makes them much more valuable in an organizational context). having the right toolkit to get there is a pretty crucial piece of the puzzle! take managed deep agents for a spin, where proactivity is baked in as first principles: langch.in/mda

Sydney Runkle @sydneyrunkle 20h Article What We’ve Learned Building Proactive Agents Think about how you work with a new hire. In their first week, you give them small, well-scoped tasks, like triaging a bug report or updating a doc. Soon you're giving them bigger tasks, and

Caspar avatar
⟳ nathan drezner retweeted
Caspar@caspar_br·21h

Running a workshop with @ishaan_lal1 today on Managed Deep Agents, come along to see what the future of agents looks like !

Happening tomorrow in SF! Join our hands-on workshop exploring Managed Deep Agents on LangSmith. You’ll get practical experience building, testing, and deploying a production-ready data analysis agent. Spaces are limited! RSVP ⤵️ fieldevents.langchain.com/bu…

nathan drezner avatar
nathan drezner@ndrezn·Nov 13

langchain just raised our series c! own. your. intelligence. with managed deep agents that means custom models, open-source harness, self-managed proactive agents, and a platform to back up everything you build. there's so much to do. can't wait to keep building 🚀

Post attachment
nathan drezner avatar
nathan drezner@ndrezn·Nov 13

want an open harness, connected to infrastructure you can monitor and manage yourself? Managed Deep Agents. It's backed by the beloved open-source Deep Agents harness and runs on the LangSmith platform. Pick your cloud, or run it on your infrastructure with BYOC.

Hey @AnthropicAI @katelyn_lesse I ❤️ the self-hosted Claude Managed Agents framework you've built. Any chance you can make it open source so that everyone can benefit running agents safely and securely? It would line up with your company values 👇

Post attachment
nathan drezner avatar
nathan drezner@ndrezn·Nov 13

hey you! 🫵 yeah you! if you're not building proactive agents yet, @ishaan_lal1's new course is the fastest way to push the pedal to the metal and build out production-grade agents. take a dive 👇

New LangChain Academy course: Building and Deploying Managed Agents 🚀 An agent can take weeks to ship to real users because of all the infrastructure that has to run alongside it. Led by @ishaan_lal1, learn to close that runtime gap Enroll free: academy.langchain.com/course…

1:39
LangChain OSS avatar
⟳ nathan drezner retweeted
LangChain OSS@LangChain_OSS·Nov 11

LangChain has day zero support for Google's new Decisions API! Gemini + TypeSafe classifiers share a base class, so middleware takes either.

Post attachment
Harrison Chase avatar
⟳ nathan drezner retweeted
Harrison Chase@hwchase17·Nov 11

we're running our annual "State of Agent Engineering" survey! we're all just trying to figure out how to build agents together - these are a few of the questions/topics that are top of mind of us excited to see what the community thinks! form.typeform.com/to/U2EAO1L…

Post attachment
nathan drezner avatar
nathan drezner@ndrezn·Nov 9

if you'll be in SF next week be sure to stop by the office for a great session on Managed Deep Agents

You're invited to learn how to deploy a production-ready agent in one evening using Managed Deep Agents. 🗓️ Nov 17th | 5:30-8:30 📍LangChain SF Office 🎟️ Info + RSVP: fieldevents.langchain.com/bu…

Post attachment
nathan drezner avatar
nathan drezner@ndrezn·Nov 5

icymi - browser use in managed deep agents! it is a wicked great experience

Stagehand @Stagehanddev Nov 5 Article Introducing Stagehand 4.4: browser.disconnect(), crash-safe sessions, and more We're excited to announce Stagehand 4.4! This release is about long-lived browsers. You can now let go of a browser without closing it, come back to it later, and trust Stagehand to clean up when your

nathan drezner avatar
nathan drezner@ndrezn·Nov 5

async subagents mean an orchestrator can pass on context to a fleet (swarm!) of subagents in order to handle many tasks in parallel, without blocking the original interaction point. locally your cc instance does this all the time - but in production it's a nightmare to manage. managing passing down context, picking up threads when a subagent completes, running parallel processes... lots to deal with. with managed deep agents it's a single configuration point in your agent definition, I'm really excited with how the devex on this feature came out.

Sydney Runkle @sydneyrunkle Nov 5 Article Agents Are Going Async Imagine a chief of staff who disappears every time you hand them a big project and only comes back once it's done. You'd probably fire them immediately. A good chief of staff keeps projects moving in

Sydney Runkle avatar
⟳ nathan drezner retweeted
Sydney Runkle@sydneyrunkle·Nov 4

if you give your agent a browser, it can do work on the web, even on sites w/ no agent native APIs! in this demo my agent books dinner reservations for me. i only had to log into the reservation site once; it saved my login and acts on my behalf for every new session!

5:33
LangChain avatar
⟳ nathan drezner retweeted
LangChain@LangChain·Nov 2

🌉 Interrupt, The Agent Conference by LangChain, returns to San Francisco March 24 + 25 🎟️ Early Bird tickets available until Dec 7th Info + RSVP: interrupt.langchain.com/

Post attachment
Josh Rosen avatar
⟳ nathan drezner retweeted
Josh Rosen@JoshARosen·Oct 29

“Managed Agents” are infra. “Model Routing” is infra. “Sandboxes” are infra. “Orchestration” is infra. “Multi-agent composition” is infra. It’s all infra. The question is simply going to be whether you choose neutral infrastructure providers or lock into the labs’ infrastructure.