# AgentSky — Use Cases https://agentsky.dev/use-cases ## Agent Arena Canonical URL: https://agentsky.dev/use-cases/evaluate-agent-stacks See which agent wins on your actual work. AgentSky's Agent Playground has run 40,000 tasks across 10 agent and model combinations in 10 categories. Pick the configuration that performs best on work like yours — then start a compatible cloud agent from the same screen. ### Choose AgentSky when Teams using public, vendor-observed benchmark runs to shortlist supported harness-and-model pairs before they invest in a task-specific evaluation. ### Choose another approach when Run your own task-level evaluation when proprietary scorers, restricted data, custom environments, or unsupported stacks must determine the winner. ### What you choose - Task category - Agent stacks to compare - Quality vs. cost bar ### What AgentSky runs - A cloud sandbox per agent - Side-by-side cost and time tracking - One API across every stack ### Evidence Evidence type: Agent Playground 40,000+ tasks. 10 agent stacks. 10 categories. Rankings come from live cloud runs — not benchmarks written for the occasion. Evidence scope: AgentSky observational production data and your own comparison runs; not controlled proof of future or universal performance. Evidence link: https://agentsky.dev/benchmarks Evidence link label: Inspect the evidence ### Next step Primary: [Open Agent Playground](https://agentsky.dev/benchmarks) Alternative: [Try an agent](https://agentsky.dev/playground/enter) ## Overnight migrations Canonical URL: https://agentsky.dev/use-cases/cloud-coding-agent Queue the migration. Wake up to a pull request. AgentSky runs migrations, refactors, and multi-step research on a persistent cloud computer. The same runtime that has executed 40,835 benchmarked tasks keeps the session available beyond your laptop connection and recovers supported interruptions without manual runtime operations. ### Choose AgentSky when Engineering teams delegating multi-step repository work that must outlast a laptop, network connection, or one teammate's availability. ### Choose another approach when Keep the agent local when code cannot leave the developer machine, the job needs unsupported local hardware or networking, or the work is short and fully interactive. ### What you choose - Repository and task - Agent harness and model - Instructions and tools ### What AgentSky runs - Persistent cloud computer - Files, state, and session history - Recovery through the night ### Evidence Evidence type: Developer architecture Tycoon runs multi-hour coding tasks on AgentSky and resumes them from Slack or the API — no persistent server to operate, no session to restart after a pause. Evidence scope: AgentSky's architecture account using Tycoon implementation context; not an independent uptime, throughput, or savings study. Evidence link: https://agentsky.dev/case-studies/tycoon Evidence link label: Read the architecture account ### Next step Primary: [Launch an agent](https://agentsky.dev/playground/enter?source=use-case-cloud-coding) Alternative: [Read the developer API](https://agentsky.dev/developers) ## Code review workflows Canonical URL: https://agentsky.dev/use-cases/code-review Put an agent behind every review — without operating a fleet. AgentSky starts one isolated cloud agent per review job, streams structured events to your product through a single API contract, and keeps that contract stable as harnesses and models change underneath. Tested across 40,835 runs and 10 agent stacks. ### Choose AgentSky when Product teams embedding an isolated review agent per job while keeping one API and event contract across supported harness and model changes. ### Choose another approach when Use deterministic CI analysis for fixed rules, or own the runtime when reviews require air-gapped deployment, custom kernels, or unsupported execution controls. ### What you choose - Review policy and tools - Agent harness and model - Repositories and permissions ### What AgentSky runs - An isolated sandbox per review - Streaming event contract - Stack upgrades without rebuild ### Evidence Evidence type: Developer case study Tycoon's code review workflow runs through AgentSky's Agent API. Tycoon configures the policy and owns the output; AgentSky provides an isolated runtime, session state, and event stream per review job. Evidence scope: AgentSky's first-party architecture account of the Tycoon integration; not an independent measure of review accuracy or business impact. Evidence link: https://agentsky.dev/case-studies/tycoon Evidence link label: Read the architecture account ### Next step Primary: [Build with the API](https://agentsky.dev/developers) Alternative: [Sign up](https://agentsky.dev/auth/sign-up) ## Agent evals Canonical URL: https://agentsky.dev/use-cases/custom-agent-skills Run your skill on every agent stack. See which wins. AgentSky runs the same skill against multiple cloud agent stacks simultaneously. Compare quality, time, and cost under identical task conditions — real runs across 10 agent and model combinations, not gut feel. ### Choose AgentSky when Teams with a fixed skill and representative task that want to run that exact workload across supported cloud agent stacks before choosing a default. ### Choose another approach when Use your own eval system when private scorers, bespoke environments, unsupported stacks, or a controlled experimental design must remain inside your infrastructure. ### What you choose - Skill and instructions - Agent stacks to evaluate - Evaluation task ### What AgentSky runs - A cloud agent per stack - Tools and connections per run - Results and cost per configuration ### Evidence Evidence type: Task-level eval workflow Agent Playground ranks 10 agent stacks across 10 categories, then lets teams run one fixed skill and task on the shortlisted stacks before choosing a default. Evidence scope: AgentSky observational rankings and task-level comparisons; not a guarantee that the leading stack will win on every skill or future workload. Evidence link: https://agentsky.dev/benchmarks Evidence link label: Inspect the evidence ### Next step Primary: [Browse skills](https://agentsky.dev/features/skills) Alternative: [Explore Agent Playground](https://agentsky.dev/benchmarks) ## Background agents Canonical URL: https://agentsky.dev/use-cases/background-agents Close the laptop. The agent keeps working. AgentSky runs your agent in a persistent cloud sandbox that retains files, state, and session history across supported pause, restart, and image-update flows. This runtime has completed 40,835 tasks across 10 agent stacks without a local machine to keep alive. Any authorized teammate can resume from web, Slack, Telegram, or the API. ### Choose AgentSky when Teams handing off long-running work across people or supported channels while the agent's cloud computer and session stay available. ### Choose another approach when Use a local agent for short device-bound work, or a self-hosted runtime when infrastructure ownership, data residency, or custom images are mandatory. ### What you choose - Task and instructions - Agent harness and model - Which channel to resume from ### What AgentSky runs - Persistent cloud sandbox - Files, state, and history - Recovery on restart ### Evidence Evidence type: Agent Hosting Teams bring their existing agent and model subscription to AgentSky. The persistent sandbox, automatic recovery, and session history run underneath — no fleet to operate, no sessions to restart manually. Evidence scope: A product capability description, not a published SLA or independent reliability benchmark; validate recovery requirements for your workload. Evidence link: https://agentsky.dev/developers Evidence link label: See capability details ### Next step Primary: [Try in Playground](https://agentsky.dev/playground/enter) Alternative: [Read the developer API](https://agentsky.dev/developers) ## Agents in Slack Canonical URL: https://agentsky.dev/use-cases/slack-agent Give your team an agent they can reach from Slack. AgentSky connects a cloud agent to Slack, Telegram, or Discord as a native channel — 3 platforms, one shared persistent session per agent. Each teammate reaches the agent from the conversation tool they already use — same session, same history, no separate interface to learn. ### Choose AgentSky when Teams that want one persistent hosted agent reachable from supported conversation channels without asking every teammate to adopt a developer tool. ### Choose another approach when Build a native bot when the job is stateless, needs an unsupported channel, or requires bespoke identity, moderation, and message-routing controls. ### What you choose - Which channels to connect - Agent and model - Instructions and tools ### What AgentSky runs - Channel connections across platforms - One shared persistent session - Sandbox and state across channels ### Evidence Evidence type: Background agents A Slack, Telegram, or Discord channel connects to the same persistent cloud session. The agent carries on between messages — close the channel and it keeps working until you resume. Evidence scope: An AgentSky product workflow linked to the background-agent capability; not independent evidence of adoption, response quality, or channel uptime. Evidence link: https://agentsky.dev/docs/channels Evidence link label: See capability details ### Next step Primary: [Launch an agent](https://agentsky.dev/playground/enter?source=use-case-slack) Alternative: [See tutorials](https://agentsky.dev/tutorials) ## Agent fleet Canonical URL: https://agentsky.dev/use-cases/run-agents-at-scale Run an agent fleet without building one. AgentSky creates one isolated cloud agent per task through a single API call. Each sandbox starts independently, runs in parallel, and streams structured results — the same approach behind 40,835 tasks run across 10 agent stacks. You keep queue and output policy; AgentSky operates the sandboxes. ### Choose AgentSky when Product teams starting many isolated agent tasks through one API while retaining control of task routing, budgets, and downstream outputs. ### Choose another approach when Own the execution layer when you need self-hosting, custom queue placement, GPU training, kernel-level controls, or workload guarantees outside AgentSky's contract. ### What you choose - Task list and instructions - Agent stack per task - Concurrency and budget ### What AgentSky runs - A cloud sandbox per session - Streaming events - Lifecycle and recovery across the fleet ### Evidence Evidence type: Software factories AgentSky's software factory pattern creates one isolated cloud agent per task through the Agent API. The product owns queue logic and output routing; AgentSky operates every sandbox, event stream, and recovery layer. Evidence scope: A reference architecture, not a capacity or throughput guarantee; concurrency, budget, and workload fit still require validation. Evidence link: https://agentsky.dev/software-factories Evidence link label: Read the reference architecture ### Next step Primary: [Read the developer API](https://agentsky.dev/developers) Alternative: [See the factory architecture](https://agentsky.dev/software-factories) ## Agent Builder Canonical URL: https://agentsky.dev/use-cases/build-custom-ai-agent Launch your own agent. Connect your team when you are ready. AgentSky hosts agents configured with your own instructions, tools, and model — in the browser or with sky launch. Each launched agent gets an API endpoint, and you can connect Slack, Telegram, or Discord across 10 compatible agent stacks. ### Choose AgentSky when Teams launching a configured hosted agent across web, API, and supported channels without first building the runtime and delivery layer. ### Choose another approach when Build a custom application when you require self-hosting, an unsupported harness or channel, or bespoke authentication and interface behavior. ### What you choose - Instructions and tools - Agent harness and model - Which channels to connect ### What AgentSky runs - Sandbox and persistent state - The agent's own API endpoint - Channel connections across platforms ### Evidence Evidence type: Channel setup After launch, teams can connect the same hosted agent to Slack, Telegram, or Discord through AgentSky's channel setup instead of building a separate bot runtime. Evidence scope: An AgentSky product capability description, not independent proof of customer adoption, agent quality, or business outcomes. Evidence link: https://agentsky.dev/docs/channels Evidence link label: See capability details ### Next step Primary: [Create your agent](https://agentsky.dev/playground/enter?source=use-case-builder) Alternative: [Try stacks in the Playground](https://agentsky.dev/playground/enter)