Cognition

nader dabit

Jared Palmer

<p>Cognition has launched Devin Outposts, a feature that splits Devin’s agent loop from its execution environment. Devin’s reasoning stays in Cognition’s cloud, but every command, file edit, and process runs on infrastructure you control.</p>
<p>That could be a Mac mini in your office, a GPU box in your lab, a VM inside a private network, or a Kubernetes cluster sitting next to your internal services. One user already has Devin running on a local Mac mini specifically to handle private customer data that can’t leave the building - the kind of use case that’s just not possible when everything runs in a third-party data center.</p>
<h2>Launch partners</h2>
<p>Several platforms launched integrations on day one:</p>
<ul>
<li><strong>Modal</strong> - open-source integration called <code>modal-devin</code>, supporting GPU-backed sessions, custom Docker images, workspace snapshotting, and fast cold-starts. Setup is a single command: <code>uvx modal-devin init</code>, which generates an orchestrator and worker function.</li>
<li><strong>Cloudflare</strong> - Devin agents run in isolated sandboxes orchestrated by Workers.</li>
<li><strong>Daytona</strong> - supports both Linux and Windows sandboxes, with per-team isolated outposts.</li>
<li><strong>E2B</strong> - also confirmed as a supported environment.</li>
</ul>
<h2>What this changes</h2>
<p>The core idea is straightforward: Devin’s brain is a cloud service, but the machine it works on is now yours. That separation matters for teams with compliance requirements, private data, or workloads that need specific hardware like GPUs.</p>
<p>Devin Outposts launched in alpha on July 21, 2026.</p>


Collections

Cognition has launched Devin Outposts, a new feature that allows running Devin AI coding sessions on user-controlled infrastructure — including Mac minis, GPU machines, private VMs, or Kubernetes clusters. The architecture splits responsibilities: Devin's reasoning remains cloud-based while all command execution and file edits happen locally on the user's hardware, enabling AI-assisted development without exposing internal environments to external access.

Devin Outposts lets you run Devin sessions on your own machines