{"id":16560,"date":"2026-11-17T11:04:48","date_gmt":"2026-11-17T02:04:48","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=16560"},"modified":"2026-11-17T11:04:48","modified_gmt":"2026-11-17T02:04:48","slug":"lm-studio-agent-remote-control-2","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/lm-studio-agent-remote-control-2\/","title":{"rendered":"LM Studio Announces Agent Remote Control Powered by LM Link"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Publisher<\/td>\n<td>LM Studio<\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-11-16<\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>LM Studio, an application that provides tools for running and managing open-weight models, has announced a new feature called &#8220;Agent Remote Control&#8221; in its &#8220;Bionic&#8221; tool, allowing users to execute and manage remote processing across multiple devices. This feature enables users to set any device they own\u2014such as a laptop, desktop, development machine, or virtual machine (VM) in the cloud\u2014as an agent host, allowing them to create projects, operate sessions, and continue work from another device.<\/p>\n<p>Agent Remote Control operates on &#8220;LM Link,&#8221; a communication infrastructure built through a partnership with Tailscale. This ensures that all communication for controlling remote agents and exchanging session information is end-to-end encrypted without exposing data to the public internet.<\/p>\n<h2>Announcement Details<\/h2>\n<p>According to the announcement, the main features of Agent Remote Control and its underlying technology, LM Link, are as follows:<\/p>\n<h3>Features and Operation of Agent Remote Control<\/h3>\n<p>Agent Remote Control is a feature designed to run LLM agents directly on a user&#8217;s other computing resources and securely control them from a local operating device.<\/p>\n<ul>\n<li><strong>Multi-Platform and Work Transfer<\/strong>: In cross-platform development, users can have agents on a Mac directly build or test on Windows or Linux machines. Additionally, long-running tasks such as performance optimization can be handed off to a remote machine, allowing work to continue even if the user closes their laptop screen.<\/li>\n<li><strong>Control Interface and @ Mentions<\/strong>: When creating a project, users can select the execution device, and type &#8220;@&#8221; in the session message input field to specify a particular device and dispatch tasks to a remote machine. Sessions on the dispatched remote machine display which session sent the message.<\/li>\n<li><strong>Support for Headless Environments<\/strong>: For servers or VMs without displays, users can set them up as remote control hosts by installing <code>llmster<\/code>, a dedicated headless daemon.<\/li>\n<li><strong>Device Control Permissions<\/strong>: On the machine to be remotely controlled, users enable the &#8220;Allow remote agent control&#8221; setting and create and register a device network from &#8220;Settings &gt; Devices &gt; Remote Machines&#8221; on the controlling device.<\/li>\n<\/ul>\n<h3>LM Link and Integration with Tailscale as Technical Infrastructure<\/h3>\n<p>&#8220;LM Link,&#8221; which supports the communication for Agent Remote Control, is an end-to-end encrypted transport developed to securely share and utilize open-weight LLMs in self-hosted environments.<\/p>\n<ul>\n<li><strong>Private Peer-to-Peer Connections<\/strong>: Authenticated devices automatically discover each other and establish direct connections regardless of their network environment. Because endpoints are not exposed to the public internet, connections are safely maintained.<\/li>\n<li><strong>Adoption of tsnet<\/strong>: LM Link utilizes <code>tsnet<\/code> provided by Tailscale (a user-space Go program that operates without modifying kernel sockets or routing tables). This enables quick, auditable initial setup, centralized device control, and secure encrypted traffic.<\/li>\n<li><strong>High Privacy Protection<\/strong>: All exchanges, including prompts and response inference data, model lists, and hardware information, are encrypted between devices. This data is never viewed by Tailscale or LM Studio&#8217;s backend services.<\/li>\n<li><strong>Command-Line Control<\/strong>: Setting operations can be completed in a few clicks from the GUI application, or via terminal commands (<code>lms login<\/code> and <code>lms link enable<\/code>) for enablement and device authentication.<\/li>\n<\/ul>\n<h2>Background<\/h2>\n<p>LM Link is a communication infrastructure developed through a partnership between LM Studio and Tailscale with the goal of making it easy and secure to run and share open-weight LLMs on self-managed hardware. Traditionally, LM Link was primarily utilized for purposes such as &#8220;loading large models onto the device with the highest computing power and sending remote inference requests from other terminals to use them.&#8221;<\/p>\n<p>However, it was demonstrated that this peer-to-peer encrypted transport, integrating Tailscale&#8217;s private networking technology, is extremely effective not only as a model inference call, but also as a foundation for remotely operating and managing projects and sessions of LLM agents running in remote environments. The background behind the release of &#8220;Agent Remote Control&#8221; stems from the needs of users and development teams who want to maximize their own hardware resources while securely sharing LLM processing and models across multiple terminals without exposing data to the public internet.<\/p>\n<h2>Impact on Local LLM Users<\/h2>\n<p>For engineers operating open-weight models on local PCs, self-hosted servers, and other equipment, the specifications of Agent Remote Control and LM Link have a direct impact on model deployment forms and security aspects.<\/p>\n<h3>Utilization of Hybrid Computing Resources and Local Operation<\/h3>\n<p>Flexible operations combining local terminals and remote computing resources become possible without relying on public cloud APIs. For example, a configuration can be realized where a lightweight model (Qwen3 4B in the documentation&#8217;s example) is kept on a daily-use laptop, while requests are transparently sent via encrypted communication to an ultra-large model (GPT-OSS 120B in the documentation&#8217;s example) on a machine with a powerful GPU located remotely. Furthermore, for Linux servers or cloud virtual machines (VMs) without displays, the dedicated daemon <code>llmster<\/code> can be installed, and authentication and link settings can be performed via CLI tools (<code>lms login<\/code> and <code>lms link enable<\/code>) to integrate them as headless hosts.<\/p>\n<h3>Data Privacy and Encrypted Peer-to-Peer Communication<\/h3>\n<p>LM Link uses Tailscale&#8217;s user-space Go library <code>tsnet<\/code> to establish direct connections between devices. It operates without modifying kernel sockets or routing tables, achieving quick, auditable device additions and centralized control. Exchanges such as prompts, generated responses, model lists, and hardware information are all end-to-end encrypted between devices, ensuring they are not exposed to the public internet or sent to and viewed by Tailscale or LM Studio&#8217;s backend servers. This allows development teams working in environments with strict data management restrictions or CI infrastructures to securely validate and share large models.<\/p>\n<h3>Availability and Licensing<\/h3>\n<p>LM Studio is provided free of charge not only for personal use, but also for work use. Additionally, enterprise plans are available for organizations requiring advanced operational features. While direct mention of individual license terms for existing models or future open-weight model release plans is not included in the documentation, feature enhancements are provided to improve the convenience of local model serving and agent execution on self-hosted infrastructure.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/20\/inco-ai-releases-splash-engine-for-mac\/\">Inco AI Releases Splash Engine for Fast Local LLMs on Mac<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/18\/lm-studio-bionic-introspection-session-references\/\">LM Studio Announces Session References and Introspection in Bionic<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/lmstudio.ai\/blog\/agent-remote-control\">Introducing Agent Remote Control<\/a><\/li>\n<li><a href=\"https:\/\/tailscale.com\/blog\/lm-link-remote-llm-access\">LM Link: Use local models on remote devices, powered by Tailscale<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>LM Studio introduces Agent Remote Control for Bionic, enabling secure, encrypted remote LLM agent execution across multiple devices via LM Link.<\/p>\n","protected":false},"author":1,"featured_media":16559,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[105],"tags":[5156,163,5158,1523,5160,1547],"class_list":["post-16560","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-engines-and-tools","tag-agent-remote-control-en","tag-gguf-en","tag-lm-link-en","tag-lm-studio-en","tag-tailscale-en","tag-verified"],"lang":"en","translations":{"en":16560,"ja":16558},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16560","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=16560"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16560\/revisions"}],"predecessor-version":[{"id":17218,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/16560\/revisions\/17218"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/16559"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=16560"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=16560"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=16560"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}