{"id":32528,"date":"2026-07-29T13:07:42","date_gmt":"2026-07-29T11:07:42","guid":{"rendered":"https:\/\/contabo.com\/blog\/?p=32528"},"modified":"2026-07-29T13:07:45","modified_gmt":"2026-07-29T11:07:45","slug":"vps-specs-for-ai-agents","status":"publish","type":"post","link":"https:\/\/contabo.com\/blog\/vps-specs-for-ai-agents\/","title":{"rendered":"VPS Sizing Guide for Self-Hosted AI Agents in 2026"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">The answer depends on one question before any spec number: is your agent calling an external API (OpenAI, Claude, Gemini) or running the model locally? When the agent uses external APIs, the VPS handles orchestration only \u2014 a lightweight process that needs surprisingly little hardware. A <a href=\"https:\/\/contabo.com\/en\/vps\/\">Contabo Cloud VPS <\/a>4 (4 vCPU \/ 8 GB RAM \/ NVMe) is the comfortable production baseline for most API-driven setups, including multi-agent orchestration platforms like Paperclip with up to 20 concurrent agents. Local LLM inference changes everything, and requires a separate tier of hardware.<\/p>\n\n\n\n<h2 id=\"h-ai-agent-vps-specs-at-a-glance-tier-table\" class=\"wp-block-heading\">AI Agent VPS Specs at a Glance: Tier Table<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The table below covers the five main self-hosted agent workloads, from a single API-driven agent to local LLM inference. All figures reflect the production-comfortable baseline, not the absolute minimum.<\/p>\n\n\n\n<div class=\"contabo-spec-table-wrap\">\n<style>\n.contabo-spec-table-wrap{overflow-x:auto;margin:1.5em 0;font-family:Calibri,Candara,Segoe,\"Segoe UI\",Optima,Arial,sans-serif;}\ntable.contabo-spec-table{width:100%;border-collapse:collapse;font-size:15px;line-height:1.5;background:#fff;}\ntable.contabo-spec-table caption{caption-side:top;text-align:left;font-weight:600;font-size:1.08em;color:#365F91;padding:0 2px 10px;}\ntable.contabo-spec-table thead th{background:#4F81BD;color:#fff;font-weight:600;text-align:left;padding:10px 14px;border:1px solid #3a6ea1;}\ntable.contabo-spec-table tbody th[scope=\"row\"]{background:#EAF0F8;color:#1F3B57;font-weight:600;text-align:left;padding:10px 14px;border:1px solid #d7e2f0;white-space:nowrap;}\ntable.contabo-spec-table tbody td{padding:10px 14px;border:1px solid #d7e2f0;vertical-align:top;}\ntable.contabo-spec-table tbody tr:nth-child(even) td,\ntable.contabo-spec-table tbody tr:nth-child(even) th[scope=\"row\"]{background:#f7fafd;}\n.contabo-spec-table .tag{display:inline-block;padding:2px 9px;border-radius:10px;font-size:.85em;font-weight:600;}\n.contabo-spec-table .tag-no{background:#eceff1;color:#607d8b;}\n.contabo-spec-table .tag-optional{background:#fff3cd;color:#8a6d1a;}\n.contabo-spec-table .tag-yes{background:#e1f3e3;color:#1e7d32;}\n.contabo-spec-table-note{font-size:.85em;color:#5b6b7c;margin-top:8px;font-family:Calibri,Arial,sans-serif;}\n<\/style>\n<table class=\"contabo-spec-table\">\n<caption>AI Agent VPS Specs at a Glance (2026): Sizing Tiers by Workload<\/caption>\n<thead>\n<tr>\n<th scope=\"col\">Workload<\/th>\n<th scope=\"col\">vCPU<\/th>\n<th scope=\"col\">RAM<\/th>\n<th scope=\"col\">Storage<\/th>\n<th scope=\"col\">GPU<\/th>\n<th scope=\"col\">Contabo Product<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr><th scope=\"row\">API-driven single agent (Hermes, OpenClaw, n8n)<\/th><td>2 vCPU<\/td><td>4 GB<\/td><td>40 GB NVMe<\/td><td><span class=\"tag tag-no\">No<\/span><\/td><td>Cloud VPS 4<\/td><\/tr>\n<tr><th scope=\"row\">Multi-agent orchestration, API-driven, \u226420 agents (Paperclip, CrewAI, LangGraph)<\/th><td>4 vCPU<\/td><td>8 GB<\/td><td>80 GB NVMe<\/td><td><span class=\"tag tag-no\">No<\/span><\/td><td>Cloud VPS 4<\/td><\/tr>\n<tr><th scope=\"row\">Browser automation stack (Playwright\/Selenium + agents)<\/th><td>4 vCPU<\/td><td>16 GB<\/td><td>100 GB NVMe<\/td><td><span class=\"tag tag-no\">No<\/span><\/td><td>Cloud VPS 6<\/td><\/tr>\n<tr><th scope=\"row\">RAG + vector database + agents (Qdrant + multi-agent)<\/th><td>4\u20138 vCPU<\/td><td>16 GB<\/td><td>200 GB NVMe<\/td><td><span class=\"tag tag-no\">No<\/span><\/td><td>Cloud VPS 6<\/td><\/tr>\n<tr><th scope=\"row\">Local LLM, 7B model (Ollama + agents)<\/th><td>8 vCPU<\/td><td>32 GB<\/td><td>200 GB NVMe<\/td><td><span class=\"tag tag-optional\">Optional<\/span><\/td><td>Cloud VPS 8<\/td><\/tr>\n<tr><th scope=\"row\">Local LLM, 13B+ model \/ production AI SDR<\/th><td>8+ vCPU<\/td><td>64 GB+<\/td><td>500 GB+ NVMe<\/td><td><span class=\"tag tag-yes\">Yes (16\u201348 GB VRAM)<\/span><\/td><td>Cloud VDS or GPU VPS<\/td><\/tr>\n<\/tbody>\n<\/table>\n<p class=\"contabo-spec-table-note\">Figures reflect production-comfortable baselines, not absolute minimums. Verify current pricing and specs at contabo.com before provisioning.<\/p>\n<\/div>\n\n\n\n<h2 class=\"wp-block-heading\">The Rule That Governs Everything: API vs Local Inference<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">When an agent calls OpenAI, Claude, Gemini, or a similar external API, the VPS runs only the orchestration layer: a Node.js or Python process, a Postgres or SQLite database, some HTTP routing, and possibly a few Docker containers for agent sandboxes. The heavy compute \u2014 the actual language model inference \u2014 happens on the API provider&#8217;s GPU cluster, not on your server. The result is that an API-driven agent is surprisingly lightweight to host.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Running a local LLM (via Ollama, vLLM, or llama.cpp) changes this entirely. The VPS becomes the inference server, and RAM requirements jump from single-digit gigabytes to 16\u201364 GB depending on the model size. Most self-hosted agent deployments in 2026 are API-driven \u2014 understanding this split is the most important step before sizing any server.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">API-Driven Agents: Paperclip, Hermes, n8n, CrewAI<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">For the majority of self-hosted agent setups in 2026, the VPS is running: an orchestration process (Node.js for <a href=\"https:\/\/contabo.com\/en\/paperclip-ai-vps-hosting\/\">Paperclip<\/a>, Python for CrewAI\/LangGraph, or a workflow engine like <a href=\"https:\/\/contabo.com\/en\/n8n-hosting\/\">n8n<\/a>), a Postgres or SQLite <a href=\"https:\/\/contabo.com\/en\/database-hosting\/\">database<\/a> for state and audit logs, a web dashboard, and optionally Docker containers providing isolated sandboxes per agent.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Paperclip&#8217;s own documentation lists a minimum of 1 vCPU and 1\u20132 GB RAM, with 2 vCPU and 2 GB as sufficient to get started. The production-comfortable baseline \u2014 handling ~20 concurrent agents with PostgreSQL, Docker sandboxes, and an active web dashboard \u2014 is 4 vCPU and 8 GB RAM. <a href=\"https:\/\/contabo.com\/de\/hermes-ai-vps-hosting\/\">Hermes Agent<\/a> has an even lower minimum: 1 vCPU and 1 GB RAM when routing all inference to external APIs. Adding n8n as a workflow automation layer alongside agents adds approximately 512 MB to 1 GB of RAM overhead but keeps the total within the 8 GB tier for most workloads. Contabo&#8217;s Cloud VPS 4 (4 vCPU \/ 8 GB \/ NVMe SSD) covers all of these configurations with comfortable headroom.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Browser Automation Adds Significant RAM: Playwright and Selenium Stacks<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">One of the most common causes of spec underestimation in AI agent deployments is underestimating how much RAM browser automation consumes. When an agent autonomously browses the web \u2014 doing lead research, scraping pages, filling forms, or running Playwright-based workflows \u2014 each headless Chromium instance uses 1\u20132 GB of RAM on its own. An agent stack that starts comfortably within 8 GB can exceed that limit quickly once browser automation enters the picture.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The practical minimum for a browser-automation agent stack is 16 GB of RAM: 8 GB for the orchestration layer, database, and application overhead, plus 4\u20138 GB reserved for concurrent browser instances. Contabo&#8217;s Cloud VPS 6 (8 vCPU \/ 16 GB) is the right tier for any agent setup that includes Playwright, Selenium, or similar web automation tooling.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">RAG Pipelines and Vector Databases<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Retrieval-augmented generation (RAG) \u2014 connecting agents to a vector database so they can search a private knowledge base before responding \u2014 adds RAM and storage requirements on top of the base orchestration spec.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Qdrant, ChromaDB, and Weaviate are the most common vector stores in self-hosted agent stacks. A mid-size collection (100,000\u20131 million vectors) typically requires 2\u20134 GB of additional RAM and 50\u2013100 GB of extra storage for the index and embeddings. A full RAG stack \u2014 agents plus vector database plus embedding pipeline \u2014 is comfortable at 16 GB RAM and 200 GB NVMe. Note: the embedding generation step (running the embedding model itself) adds further load; using an external embedding API (OpenAI, Voyage AI, Cohere) avoids this cost at the price of sending document text to a third party.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Local LLM Inference: When You Stop Sending Data to External APIs<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Running a language model locally \u2014 most commonly via <a href=\"https:\/\/contabo.com\/blog\/what-is-ollama-and-how-to-use-it-with-n8n\/\">Ollama<\/a> \u2014 is the right choice when data privacy is the primary concern: sensitive customer data, regulated industries, or compliance requirements that prevent data from leaving your infrastructure. It is also the choice when per-token API costs become significant at scale.<\/p>\n\n\n\n<div class=\"contabo-spec-table-wrap\">\n<style>\n.contabo-spec-table-wrap{overflow-x:auto;margin:1.5em 0;font-family:Calibri,Candara,Segoe,\"Segoe UI\",Optima,Arial,sans-serif;}\ntable.contabo-spec-table{width:100%;border-collapse:collapse;font-size:15px;line-height:1.5;background:#fff;}\ntable.contabo-spec-table caption{caption-side:top;text-align:left;font-weight:600;font-size:1.08em;color:#365F91;padding:0 2px 10px;}\ntable.contabo-spec-table thead th{background:#4F81BD;color:#fff;font-weight:600;text-align:left;padding:10px 14px;border:1px solid #3a6ea1;}\ntable.contabo-spec-table tbody th[scope=\"row\"]{background:#EAF0F8;color:#1F3B57;font-weight:600;text-align:left;padding:10px 14px;border:1px solid #d7e2f0;white-space:nowrap;}\ntable.contabo-spec-table tbody td{padding:10px 14px;border:1px solid #d7e2f0;vertical-align:top;}\ntable.contabo-spec-table tbody tr:nth-child(even) td,\ntable.contabo-spec-table tbody tr:nth-child(even) th[scope=\"row\"]{background:#f7fafd;}\n.contabo-spec-table .tag{display:inline-block;padding:2px 9px;border-radius:10px;font-size:.85em;font-weight:600;}\n.contabo-spec-table .tag-optional{background:#fff3cd;color:#8a6d1a;}\n.contabo-spec-table .tag-recommended{background:#fde3cf;color:#a1500a;}\n.contabo-spec-table .tag-required{background:#fbe0e0;color:#c0392b;}\n.contabo-spec-table-note{font-size:.85em;color:#5b6b7c;margin-top:8px;font-family:Calibri,Arial,sans-serif;}\n<\/style>\n<table class=\"contabo-spec-table\">\n<caption>Local LLM Inference in 2026: RAM and GPU Requirements by Model Size<\/caption>\n<thead>\n<tr>\n<th scope=\"col\">Model Size<\/th>\n<th scope=\"col\">RAM Needed<\/th>\n<th scope=\"col\">Storage<\/th>\n<th scope=\"col\">GPU<\/th>\n<th scope=\"col\">Notes<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr><th scope=\"row\">7B model (Hermes 3 7B, Llama 3.1 8B)<\/th><td>16\u201332 GB<\/td><td>200 GB NVMe<\/td><td><span class=\"tag tag-optional\">Optional<\/span><\/td><td>CPU inference works but is slow; fine for batch\/non-real-time tasks.<\/td><\/tr>\n<tr><th scope=\"row\">13B model<\/th><td>32\u201364 GB<\/td><td>300 GB NVMe<\/td><td><span class=\"tag tag-recommended\">Recommended<\/span><\/td><td>CPU inference impractical for production use.<\/td><\/tr>\n<tr><th scope=\"row\">34B+ model<\/th><td>64 GB+<\/td><td>500 GB+ NVMe<\/td><td><span class=\"tag tag-required\">Required (16\u201348 GB VRAM)<\/span><\/td><td>Use Contabo GPU VPS or Cloud VDS.<\/td><\/tr>\n<\/tbody>\n<\/table>\n<p class=\"contabo-spec-table-note\">Grounded in Paperclip\/Hermes official documentation and independent AI-response consensus on model hardware requirements.<\/p>\n<\/div>\n\n\n\n<p class=\"wp-block-paragraph\">One important nuance: if data privacy is the reason for self-hosting but you don&#8217;t need local inference, hosting an API-driven agent on a Contabo VPS in the <a href=\"https:\/\/contabo.com\/en\/locations\/europe\/\">EU Hub <\/a>(Lauterbourg) keeps your orchestration layer, agent state, and customer data within EU infrastructure under German-owned hosting \u2014 even if the LLM calls go to Anthropic or OpenAI. This is a meaningful GDPR data-residency position that doesn&#8217;t require local inference and doesn&#8217;t require a 32 GB server.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Mapping Each Tier to a Contabo Product<\/h2>\n\n\n\n<div class=\"contabo-spec-table-wrap\">\n<style>\n.contabo-spec-table-wrap{overflow-x:auto;margin:1.5em 0;font-family:Calibri,Candara,Segoe,\"Segoe UI\",Optima,Arial,sans-serif;}\ntable.contabo-spec-table{width:100%;border-collapse:collapse;font-size:15px;line-height:1.5;background:#fff;}\ntable.contabo-spec-table caption{caption-side:top;text-align:left;font-weight:600;font-size:1.08em;color:#365F91;padding:0 2px 10px;}\ntable.contabo-spec-table thead th{background:#4F81BD;color:#fff;font-weight:600;text-align:left;padding:10px 14px;border:1px solid #3a6ea1;}\ntable.contabo-spec-table tbody th[scope=\"row\"]{background:#EAF0F8;color:#1F3B57;font-weight:600;text-align:left;padding:10px 14px;border:1px solid #d7e2f0;white-space:nowrap;}\ntable.contabo-spec-table tbody td{padding:10px 14px;border:1px solid #d7e2f0;vertical-align:top;}\ntable.contabo-spec-table tbody tr:nth-child(even) td,\ntable.contabo-spec-table tbody tr:nth-child(even) th[scope=\"row\"]{background:#f7fafd;}\n.contabo-spec-table-note{font-size:.85em;color:#5b6b7c;margin-top:8px;font-family:Calibri,Arial,sans-serif;}\n<\/style>\n<table class=\"contabo-spec-table\">\n<caption>2026 Sizing Tiers Mapped to Contabo Products for Self-Hosted AI Agents<\/caption>\n<thead>\n<tr>\n<th scope=\"col\">Tier<\/th>\n<th scope=\"col\">Typical Workload<\/th>\n<th scope=\"col\">Contabo Product<\/th>\n<th scope=\"col\">Notes<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr><th scope=\"row\">Entry<\/th><td>API-driven single agent (Hermes, OpenClaw, simple n8n)<\/td><td>Cloud VPS 4 \u2014 4 vCPU \/ 8 GB \/ NVMe<\/td><td>Exceeds Paperclip and Hermes minimum; plenty of headroom for growth.<\/td><\/tr>\n<tr><th scope=\"row\">Production API<\/th><td>Multi-agent \u226420, Paperclip, CrewAI, n8n + agents<\/td><td>Cloud VPS 4 \u2014 4 vCPU \/ 8 GB \/ NVMe<\/td><td>Comfortable for continuous production workloads with Postgres and Docker.<\/td><\/tr>\n<tr><th scope=\"row\">Extended<\/th><td>Browser automation (Playwright) or RAG + vector DB<\/td><td>Cloud VPS 6 \u2014 8 vCPU \/ 16 GB \/ NVMe<\/td><td>16 GB covers headless browser instances and Qdrant\/ChromaDB overhead.<\/td><\/tr>\n<tr><th scope=\"row\">Local LLM (small)<\/th><td>Ollama 7B alongside agents<\/td><td>Cloud VPS 8 \u2014 24 GB RAM<\/td><td>CPU inference; practical for batch\/async tasks, not real-time chat.<\/td><\/tr>\n<tr><th scope=\"row\">Local LLM (production)<\/th><td>13B+ models, heavy AI SDR, Hermes 4<\/td><td>Cloud VDS or GPU VPS<\/td><td>Dedicated CPU + GPU option; verify current GPU VPS availability at contabo.com.<\/td><\/tr>\n<\/tbody>\n<\/table>\n<p class=\"contabo-spec-table-note\">Product names and specs evolve \u2014 confirm current lineup and pricing at contabo.com before publishing.<\/p>\n<\/div>\n\n\n\n<p class=\"wp-block-paragraph\">All Contabo plans include unlimited traffic with no egress billing \u2014 relevant for agent workloads that pull data from the web, stream API responses, or push large volumes of orchestration logs. Verify current pricing and exact specs at contabo.com before provisioning, as product names and specs update over time. <\/p>\n\n\n\n<div class=\"wp-block-uagb-advanced-heading uagb-block-5ce775e3\"><h2 class=\"uagb-heading-text\">FAQ: VPS Specs for AI Agents<\/h2><\/div>\n\n\n\n<div class=\"schema-faq wp-block-yoast-faq-block\"><div class=\"schema-faq-section\" id=\"faq-question-1785322708773\"><strong class=\"schema-faq-question\">What VPS specs do I need to run Paperclip with multiple agents?<\/strong> <p class=\"schema-faq-answer\">For Paperclip with multiple concurrent agents using external APIs (Claude, GPT-4, Gemini): 4 vCPU, 8 GB RAM, and 80 GB NVMe is the production-comfortable baseline for up to ~20 concurrent agents. Paperclip&#8217;s own documented minimum is 1 vCPU \/ 1\u20132 GB RAM, but in practice the PostgreSQL database, web dashboard, and Docker agent sandboxes together consume most of that before agents start working. Contabo&#8217;s Cloud VPS 4 covers this tier comfortably.<\/p> <\/div> <div class=\"schema-faq-section\" id=\"faq-question-1785322719040\"><strong class=\"schema-faq-question\">How much RAM does an AI agent need on a VPS?<\/strong> <p class=\"schema-faq-answer\">For a single API-driven agent: 2\u20134 GB is sufficient. For multi-agent orchestration with a database and web dashboard: 8 GB is the practical baseline. Add browser automation (Playwright\/Selenium) and plan for 16 GB. Add a local LLM and plan for 16\u201332 GB minimum depending on model size. The most common mistake is sizing for the agent alone and forgetting the database, containers, and any browser or model overhead.<\/p> <\/div> <div class=\"schema-faq-section\" id=\"faq-question-1785322730227\"><strong class=\"schema-faq-question\">Do I need a GPU VPS to run AI agents?<\/strong> <p class=\"schema-faq-answer\">No \u2014 for API-driven agents, a standard CPU VPS is sufficient. GPU becomes necessary when running a local LLM on the same server: 7B models can technically run on CPU (slowly), but anything 13B and above requires a GPU for practical inference speeds. If data privacy is the goal, hosting an API-driven agent on a Contabo EU VPS keeps your data in EU infrastructure without requiring local inference.<\/p> <\/div> <div class=\"schema-faq-section\" id=\"faq-question-1785322740623\"><strong class=\"schema-faq-question\">Which VPS is best for self-hosted AI agents in 2026?<\/strong> <p class=\"schema-faq-answer\">For most API-driven agent setups, Contabo&#8217;s Cloud VPS 4 (4 vCPU \/ 8 GB RAM \/ NVMe) offers the best RAM-per-euro ratio in the tier and includes unlimited traffic \u2014 relevant for agents that pull data from the web or push high volumes of API traffic. For browser automation or RAG workloads, Cloud VPS 6 (8 vCPU \/ 16 GB) is the right step up. For local LLM inference, Cloud VPS 8 (24 GB) handles 7B models, while 13B+ requires the Cloud VDS or GPU VPS line.<\/p> <\/div> <\/div>\n","protected":false},"excerpt":{"rendered":"<p>The answer depends on one question before any spec number: is your agent calling an external API (OpenAI, Claude, Gemini) or running the model locally? When the agent uses external APIs, the VPS handles orchestration only \u2014 a lightweight process that needs surprisingly little hardware. A Contabo Cloud VPS 4 (4 vCPU \/ 8 GB [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"_uag_custom_page_level_css":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"set","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":"","_members_access_role":[],"_members_access_error":""},"categories":[4708],"tags":[],"ppma_author":[4285],"class_list":["post-32528","post","type-post","status-publish","format-standard","hentry","category-decision-guide"],"uagb_featured_image_src":{"full":false,"thumbnail":false,"medium":false,"medium_large":false,"large":false,"1536x1536":false,"2048x2048":false},"uagb_author_info":{"display_name":"Jie Guo","author_link":"https:\/\/contabo.com\/blog\/author\/jieguo\/"},"uagb_comment_info":0,"uagb_excerpt":"The answer depends on one question before any spec number: is your agent calling an external API (OpenAI, Claude, Gemini) or running the model locally? When the agent uses external APIs, the VPS handles orchestration only \u2014 a lightweight process that needs surprisingly little hardware. A Contabo Cloud VPS 4 (4 vCPU \/ 8 GB&hellip;","authors":[{"term_id":4285,"user_id":78,"is_guest":0,"slug":"jieguo","display_name":"Jie Guo","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/4e0d981b06988d6d456834e9d55bc9e713e918fa8444325543d14f448154106b?s=96&d=mm&r=g","author_category":"","user_url":"","last_name":"Guo","first_name":"Jie","job_title":"","description":""}],"_links":{"self":[{"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/posts\/32528","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/users\/78"}],"replies":[{"embeddable":true,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/comments?post=32528"}],"version-history":[{"count":1,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/posts\/32528\/revisions"}],"predecessor-version":[{"id":32529,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/posts\/32528\/revisions\/32529"}],"wp:attachment":[{"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/media?parent=32528"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/categories?post=32528"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/tags?post=32528"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/contabo.com\/blog\/wp-json\/wp\/v2\/ppma_author?post=32528"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}