LangChain open-sources the design agent that runs its own marketing

What it does Content Studio is the design agent LangChain uses internally, now Apache 2.0. You mention it in a Slack thread with a brief and attachments, and it returns finished, on-brand files — a resized ad, a deck, a chart, or a launch film. Since August it has shipped over 3,000 assets for LangChain, and three in four requests come from outside marketing.

The interesting bit Everything it produces is code in a per-thread sandbox: a slide is HTML, a film is an animation timeline. So a revision is an edit and a re-render, not a restart. It also renders and looks at its own output before posting, and checks colors, fonts, contrast and chart math in code on every render — taste stays with your designers.

Key highlights

• Built on Deep Agents, deployed via Managed Deep Agents, which handles Slack, sandboxes, memory, schedules and tracing.
• 35 skills loaded on demand via a router skill; a design wiki holds the craft, your brand pack holds the facts.
• Async subagents split big jobs (films go storyboard → animatic → production) on a shared sandbox.
• Two-layer memory: team rules everyone shares, plus personal memory per person — tell it a rule once and it sticks.
• learn_from_job writes post-job notes that later jobs read and rate; notes that don’t hold up get dropped.

Caveats

• Firmly coupled to the LangChain stack: you need a LangSmith account (Managed Deep Agents is in public beta) and a provider model key; this is a Managed Deep Agents project, not a self-contained agent.
• The repo description says JavaScript, but the README and tooling (uv, Python 3.11+) point to Python — worth reconciling before you fork.

Verdict A genuinely interesting reference architecture for anyone building production agents with review loops, sandboxes and skill routing — and a real option if your team already lives in LangSmith. If you wanted a vendor-neutral, self-hosted design bot, this isn’t it.