AI agent hosting
AI agent hosting, with the agent included.
plori supplies a built-in AI agent and runs it for you. The agent works in its own persistent environment. Give it a task in chat or through your MCP client, then review its replies and the files it saves. You do not deploy your own agent framework or MCP server here. To host an agent you built, read the comparison guide for sandbox APIs and servers you manage.
Opens your agent straight away, no signup needed.
What do you get with a hosted AI agent?
§01A built-in agent, operated by plori
plori runs the agent and its isolated environment. The agent can run shell commands, research a question, and save a report or changed code for you to review.
Files that persist
The agent keeps files on its persistent disk and saves memory notes. Memory notes are saved context, not a snapshot of running processes.
Sleep when idle
A sleeping agent has no compute charge. Disk within your plan has no extra storage charge. Your subscription, extra disk, workflow runs, and published web services have their own costs.
Unattended work
Scheduled runs and background tasks report back into the conversation.
Approval and input requests
When the agent needs an approval or a secret, the run waits for your answer. You approve in the dashboard; an MCP client can deny a request or answer a question. Then the run continues.
One MCP connection
Connect through MCP with OAuth or an API key to create agents, send tasks, read results, and schedule work.
To choose between a local agent and a remote one, read the cloud AI agent field guide. For one bounded task from prompt to result, see the example of SEO research for a page optimization brief.
Is this hosting for my own agent?
§02No. plori provides a managed AI agent and hosts only that agent. It does not host an agent framework, agent server, or MCP server that you bring. The built-in agent can install tools and run scripts for a task, but plori does not operate your runtime.
Here is what plori does host:
- Its built-in agent and the isolated environment that the agent runs in.
- Your files on the agent's persistent disk, and the agent's memory notes.
- Scheduled runs, background tasks, and workflows that run while you are away.
- A small web service that the agent wrote. The service scales to zero when it gets no requests and starts again on the next request.
To deploy an agent you built, use a sandbox API or a server you manage. The comparison guide compares those options with plori. If you want AI agents for software development, the cloud coding agent page covers repositories, builds, and tests.
An existing local agent can also delegate tasks to plori. For state between tasks, read what persists. Compare all usage categories on pricing before a large job.
How does it work?
§03- 01Open plori.ai/agent, or connect an MCP client and create an agent.
- 02Describe the task and the result you want. Provide the files or account connections the job needs.
- 03Answer any approval or input requests. Background work can continue after you leave, within task and spending limits.
- 04Review the result in the conversation and the Files panel. Then ask for changes or schedule a follow-up.
Questions, answered.
§04What does AI agent hosting mean on plori?
plori supplies a built-in cloud AI agent with its own persistent environment, and plori operates both. You give the agent tasks in chat or through an external MCP client.
Can I host my own agent framework or MCP server on plori?
No. plori does not offer a deployment service for an agent framework, agent server, or MCP server that you bring. The built-in agent can install tools, run scripts, and build a small web service that you publish with scale to zero. For your own agent runtime, evaluate a sandbox API or a server you manage. The comparison guide covers that decision.
What costs should I include?
Compare model usage, active compute, paid tool calls, workflow runs, and storage beyond your included disk. plori charges usage in credits. A sleeping agent has no compute charge, but your subscription and extra disk can still cost money. BYOK moves model charges to your provider. Other usage remains chargeable. See pricing for current allowances and rates.
Can my existing agent send work to plori?
Yes. An external MCP client can create a plori agent, send work, read replies, answer questions, deny approval requests, and schedule runs that you confirm. The client connects with OAuth or an API key, depending on what it supports. Your local agent stays the caller, and the plori agent does the delegated task.
How are hosted agents isolated?
Each agent runs in an isolated environment with its own files. Other accounts cannot access that environment. Your agents draw storage capacity from your account disk allowance.
Which models do hosted agents use?
The plori router picks a model for each request from a shared pool. Every plan uses the same pool. With BYOK, a supported provider key bills model usage to your provider directly. Compute, paid tools, and other usage still draw from plori credits.
Can I try it without signing up?
Yes. Open plori.ai/agent to try the built-in agent without an email or card. Signing up later keeps the agent, its conversations, and its files. The trial has credit, disk, and time limits.