--- title: Cursor now lets you run cloud agents on your own machines description: Cursor's cloud agents can now run on infrastructure you control, including auto-scaling pools of your own machines, so agents can reach internal services or... canonical: https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Cursor now lets you run cloud agents on your own machines | daily.dev og:description: Cursor's cloud agents can now run on infrastructure you control, including auto-scaling pools of your own machines, so agents can reach internal services or... og:url: https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674 og:image: https://api.daily.dev/og/posts/LYXMvN674.png og:image:alt: Cursor now lets you run cloud agents on your own machines og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Cursor now lets you run cloud agents on your own machines **[Collections](https://daily.dev/sources/collections)** · 2 min read · 1 upvotes · 0 comments ## Summary Cursor's cloud agents can now run on infrastructure you control, including auto-scaling pools of your own machines, so agents can reach internal services or specialized hardware that Cursor's own servers can't access. The agent orchestration logic still lives on Cursor's side; you simply point it at machines you own, such as a repurposed Mac mini acting as a dedicated agent host. ## Content Cursor has opened up its cloud agents to run on customer-controlled infrastructure, rather than exclusively on Cursor's own hosted machines. The feature uses Cursor's Self-Hosted Machines API (available on the Enterprise plan) and lets organizations point agent execution at their own hardware, internal services, or third-party sandboxes, while Cursor still handles inference and planning. Two integrations are already live. **Vercel Sandbox** runs each agent request inside an isolated Firecracker microVM. Vercel Functions and Vercel Workflow act as a control plane that claims requests, provisions workers, monitors sessions, and cleans up afterward. The setup gives you a scale-to-zero worker pool, per-request isolation, durable retries on failure, and short-lived user-scoped credentials. A reference implementation is available to deploy directly to a Vercel account. **Coder Agent Relay**, launched with SpaceXAI as the first partner, targets regulated industries. Banks, defense agencies, and government institutions have largely been blocked from using AI coding tools because their security teams couldn't accept vendor-hosted execution environments. Agent Relay keeps tool-call execution, source code, and secrets inside the customer's own network. Cursor still handles the agent loop. Features include air-gapped deployment, sandboxed ephemeral task-scoped environments, policy defined at the environment level, and full audit logs of agent actions. It's currently in private preview. Coder CEO Rob Whiteley has been direct about what's actually slowing AI coding adoption in regulated industries: it's not the agents themselves, it's the deployment model. Security teams were blocking Cursor over architectural concerns, not specific vulnerabilities. The new setup lets developers keep using Cursor while compliance requirements get handled at the infrastructure layer. One data point Whiteley shared: at Coder, 1% of engineers account for 40% of token spend. His read is that this reflects uneven adoption across the team rather than a runaway cost problem, and that the more useful response is helping more developers actually use these tools rather than capping the heavy users. A few caveats worth noting for regulated industries considering Agent Relay: self-hosting execution doesn't eliminate the contractual and audit obligations that come with using a third-party AI provider. Under the EU's DORA regulation, Cursor still qualifies as an ICT third-party provider since it handles inference and planning, which means Article 30 and Article 28 obligations around audit rights and exit plans still apply. That's worth factoring in, especially given that SpaceX now owns Cursor and OpenAI has reportedly ended its model supply relationship with them. ## Questions this post answers ### Can Cursor's cloud agents run on infrastructure I control instead of Cursor's own servers? Yes, Cursor's Self-Hosted Machines API, available on the Enterprise plan, lets organizations run agent execution on their own hardware, internal services, or third-party sandboxes while Cursor still handles inference and planning. Two live integrations exist: Vercel Sandbox, which runs each request in an isolated Firecracker microVM, and Coder Agent Relay, built for regulated industries and currently in private preview. _Teams weighing self-hosted AI coding setups can track how tools like Cursor evolve on daily.dev._ ### Does self-hosting Cursor's agent execution remove the EU DORA compliance obligations for using it? No, self-hosting the execution layer does not eliminate DORA obligations. Cursor still qualifies as an ICT third-party provider under DORA because it handles inference and planning, so Article 30 and Article 28 requirements around audit rights and exit plans still apply even when tool-call execution runs inside the customer's own network. _Regulated-industry teams navigating AI vendor compliance can follow tooling changes like this on daily.dev._ ### Why is AI coding tool adoption uneven within engineering teams that already have access? Coder CEO Rob Whiteley reported that at Coder, 1% of engineers account for 40% of token spend, which he interprets as a sign of uneven adoption across the team rather than a runaway cost problem caused by heavy users. His suggested response is helping more developers actually use the tools rather than capping the top spenders. _Engineering leaders tracking AI coding tool adoption patterns can follow these trends on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 125 comments across x (as of 2026-09-13). **TL;DR:** Reaction is largely enthusiastic about being able to run Cursor's cloud agents on your own machines or auto-scaling pools, especially for compliance, internal-service access, and heavy builds, though several people push back on what 'self-hosted' actually means since orchestration and inference still live on Cursor's servers. **Sentiment:** 45% positive · 40% mixed · 15% skeptical **The case for** - Lets agents reach internal services and specialized hardware (GPUs, Macs) that were previously blocked behind VPNs or firewalls. - Seen as solving real enterprise compliance/security review concerns by keeping execution and secrets on the customer's own network. - Offloads heavy compute (builds, compiles) onto your own machines instead of burning cloud agent usage/credits. - Enables niche workflows like building iOS/macOS apps without owning a Mac by pointing agents at a provisioned Mac. **The pushback** - Several point out this isn't truly self-hosted since the agent loop, planning, and inference still run in Cursor's cloud, only tool execution moves locally. - Concern that repo/context data and tool outputs still round-trip to Cursor's control plane, so the privacy win is narrower than the framing suggests. - Some frustration that scaling team pools still requires an Enterprise plan and multiple confusing billing tiers. - Worry about credential scoping and blast radius when agent pools get access to internal/prod services. **By community** - x (mixed): Many replies are enthusiastic hype or emoji-level agreement, but a recurring thread of more substantive replies debates whether this is genuinely self-hosted given orchestration/inference still sit in the vendor's cloud, alongside practical questions about setup, licensing, and security scoping. **Hottest debate:** Whether keeping the agent orchestration/inference loop in Cursor's cloud while only moving tool execution to your own machines really counts as 'your infrastructure,' or is mostly a security-review talking point. **Open questions** - Which service account do pool machines run as, and can they see production systems? - Does this replace or differ from the existing self-hosted/usage pool feature from earlier this year? - Will bring-your-own-model or BYOK support be added for cloud agents? - How fast do idle pool machines spin down, and does that affect credit usage? - Is team pool support now available outside Enterprise plans? **Highlights** > @cursor_ai "on your infrastructure" is the line that gets this past a security review. the question that follows it: which service account do the pool machines run as, and does it get to see prod. the answer decides whether this is a pilot or a rollout. > — [chams\_builds on x · 1 comments](https://x.com/chams_builds/status/2095267665522597948) > @cursor_ai Important boundary: Cursor’s self-hosted machines move execution, not the whole agent. Cursor says inference + planning stay in its cloud; tool outputs flow back and may contain code. For security review, treat this as self-hosted execution—not a self-hosted agent. > — [ham\_zax on x · 2 comments](https://x.com/ham_zax/status/2095275489091088814) > @cursor_ai the split that matters: your infra runs the tool plane, the loop stays theirs. which means every observation still egresses to Cursor's model context — so 'on your infrastructure' isolates execution, not data. the privacy win is the sandbox, not the placement. > — [TonyJZhou on x](https://x.com/TonyJZhou/status/2095277173523271753) > @cursor_ai self-hosting the runners is a secrets problem, not a compute problem. the agent loop staying in cursor means repo context still round-trips to your control plane, you just moved where the code checks out. autoscaling machine pools is the easy half. > — [jr\_muj on x](https://x.com/jr_muj/status/2095269963011354836) > @cursor_ai so we pay for the self-hosted compute, we pay for the cloud infrastructure, but the actual agent control loop STILL stays on ur servers? worst of both worlds bro... u taking our money for compute while keeping us locked into ur proprietary orchestration loop 🗿 > — [AndAIyou on x](https://x.com/AndAIyou/status/2095273638878416945) **Source threads** - [x](https://x.com/cursor_ai/status/2095257412781396114) · 0 points · 125 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#infrastructure](https://daily.dev/tags/infrastructure) [View this post on daily.dev](https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Cursor now lets you run cloud agents on your own machines","url":"https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674"},"datePublished":"2026-09-02T21:08:13.204Z","dateModified":"2026-09-13T20:04:56.312Z","description":"Cursor's cloud agents can now run on infrastructure you control, including auto-scaling pools of your own machines, so agents can reach internal services or...","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,infrastructure","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Cursor now lets you run cloud agents on your own machines"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/cursor-now-lets-you-run-cloud-agents-on-your-own-machines-lyxmvn674#faq","mainEntity":[{"@type":"Question","name":"Can Cursor's cloud agents run on infrastructure I control instead of Cursor's own servers?","acceptedAnswer":{"@type":"Answer","text":"Yes, Cursor's Self-Hosted Machines API, available on the Enterprise plan, lets organizations run agent execution on their own hardware, internal services, or third-party sandboxes while Cursor still handles inference and planning. Two live integrations exist: Vercel Sandbox, which runs each request in an isolated Firecracker microVM, and Coder Agent Relay, built for regulated industries and currently in private preview. Teams weighing self-hosted AI coding setups can track how tools like Cursor evolve on daily.dev."}},{"@type":"Question","name":"Does self-hosting Cursor's agent execution remove the EU DORA compliance obligations for using it?","acceptedAnswer":{"@type":"Answer","text":"No, self-hosting the execution layer does not eliminate DORA obligations. Cursor still qualifies as an ICT third-party provider under DORA because it handles inference and planning, so Article 30 and Article 28 requirements around audit rights and exit plans still apply even when tool-call execution runs inside the customer's own network. Regulated-industry teams navigating AI vendor compliance can follow tooling changes like this on daily.dev."}},{"@type":"Question","name":"Why is AI coding tool adoption uneven within engineering teams that already have access?","acceptedAnswer":{"@type":"Answer","text":"Coder CEO Rob Whiteley reported that at Coder, 1% of engineers account for 40% of token spend, which he interprets as a sign of uneven adoption across the team rather than a runaway cost problem caused by heavy users. His suggested response is helping more developers actually use the tools rather than capping the top spenders. Engineering leaders tracking AI coding tool adoption patterns can follow these trends on daily.dev."}}]} ```