Ways to run an AI agent
How plori compares.
Choose who will build and operate the agent first. plori provides its own agent. Sandbox APIs and servers let you build or deploy your own runtime. Other products also offer managed cloud agents, persistence, and automation. Compare the work you can delegate and what you must manage.
Opens your agent straight away, no signup needed.
Who runs the agent, and what do you manage?
§01| Option | Who builds and operates it | Files and runtime state | Costs to include | Results, scheduling, and authorization |
|---|---|---|---|---|
| plori | Built-in agent operated by plori. Use chat or an external MCP client. | Files and saved notes persist. Sleep does not preserve arbitrary processes or RAM. | Usage credits cover models, compute, paid tools, and workflows. Include subscriptions and extra disk. | Replies in conversations and files you can review. One-shot scheduled runs, workflows, and approval requests are built in. |
| Sandbox APIs (E2B, Daytona, Modal) | The provider operates code execution infrastructure. You build the agent and its application. | Persistence is available. File retention and memory restoration depend on the vendor and lifecycle mode. | Include compute, storage, any plan fee, and your model and tool providers. | Your application connects execution output to users, schedules, credentials, and approvals. |
| A VPS you manage | You install and operate your own agent runtime on a rented server. | Disk retention follows the server and backup configuration. Process recovery is yours to configure. | Server rental, storage, models, tools, and operating time. Check whether stopped servers remain billable. | You implement result delivery, scheduling, permissions, updates, and monitoring. |
| A local agent on your laptop | Your agent process runs on your own device. | Local files remain on the device. Sleep, shutdown, or process failure can interrupt execution. | Hardware and electricity, plus any model or tool charges. | Local outputs and integrations you configure. Unattended execution depends on the device staying available. |
| Cloud agents in another product (for example, Cursor) | The vendor operates an agent with its own setup, integrations, and controls. | Remote work can continue without your laptop. File retention and runtime recovery follow the product lifecycle. | Check plan and model-usage charges, automation usage, and any connected tools. | Cursor supports cloud execution, MCP tools, reviewable outputs, and scheduled or event-triggered automations. |
A sandbox API can keep files and restore state. The application around it still decides what to run and how users receive the result. plori provides the agent together with persistent files, memory, and unattended work built in. See what plori hosts and current pricing.
Which setup fits the work?
§02- plori vs sandbox APIs: Use a sandbox API to run code inside an agent application you build. Use plori to assign work to its built-in agent.
- plori vs a VPS: Use a VPS when you need control of the runtime and can operate it. Compare that work and its full cost with a managed agent.
- plori vs a local agent: Keep work local when it needs your device or local-only data. Use remote execution when the task must continue without that device.
- plori vs another cloud agent: Try the cloud features of your existing product first. Add plori when its task workflow, saved files, and MCP delegation fit work you need to do.
Questions, answered.
§03Should I choose a managed agent or a sandbox API?
Choose a managed agent when you want to give tasks to an agent the provider operates. Choose a sandbox API when your application needs code execution and you want to build the agent around it. plori provides its built-in agent, including a shell, persistent files, saved notes, and task scheduling.
When is plori not the right choice?
Use another setup when you need to deploy your own agent framework, require on-premises execution, or need a graphical desktop. If your existing cloud agent already handles the task and its account connections, compare that experience before adding another service.
Is persistence unique to plori?
No. Daytona supports persistent files, E2B supports pause and resume with files and memory, and Modal provides filesystem snapshots and volumes. Check the lifecycle mode you will use. plori keeps files and saved memory notes, but does not promise RAM restoration or uninterrupted processes across sleep.
What should I compare in the total bill?
Include the subscription or server fee, model usage, active compute, paid tools, workflow runs, and storage. plori bills usage in credits. Sleeping stops agent compute charges. Extra disk and subscriptions can still cost money. BYOK moves model charges to your provider without removing other charges.
Can I move my work to another setup?
You can download files or ask the agent to push work to your own git remote or storage. Moving files does not migrate the built-in agent runtime, live processes, saved memory notes, or connected-account credentials.
Does an IDE agent need plori for unattended work?
Not necessarily. Cursor already provides cloud agents and scheduled or event-triggered automations. Add plori when you want to delegate a task to its built-in agent and return to its saved files and conversations. An external MCP client calls plori rather than becoming the agent hosted there.