Quickstart: start here
Try it free, give the agent a real first task, connect your terminal, and add credits only when you outgrow the free allowance.
Plori runs AI agents in cloud Workspaces. Give an agent a task and keep its files for the next run.
1. Start free
Open plori.ai and choose Try it free. Sign in with Google. Plori creates an agent in its own environment and opens its chat.
A registered Free account includes $5 of usage per month and 1 GB of account disk. When signup rewards are available, eligible new Google signups can receive up to $10 in welcome credit right away.
2. Give it a first task
Ask for a result that uses the environment and leaves something behind:
Research the latest releases of the project I name, compare the changes, and save a concise report to
work/report.md.
The agent can use its shell and web tools, install what it needs, and save the report in the Workspace files. Open Files to verify the file exists. Ask a follow-up in the same chat: it works on the same saved files. Software that a run installs outside the Workspace files, and processes that a run starts, do not continue to the next run.
3. Use your terminal
Install the CLI and sign in through its browser flow:
curl -fsSL https://plori.ai/install.sh | sh
plori login
plori workspace create mate
plori workspace run mate "Create a short project brief and save it to work/brief.md"
plori login opens the email-code page and returns to the terminal when complete. The
installer needs no sudo and no Node. npm i -g @plori/cli installs the same binary if
you prefer npm. It installs to ~/.local/bin and does not touch your shell config, so if
that directory is not on your PATH yet, add it first with the line the installer prints.
plori workspace create mate creates a Workspace and its agent. plori workspace run mate
sends that agent a task on the Workspace files. Creation prints an idempotency key.
Reuse that key with --idempotency-key when you retry the same creation request.
For a conversation in the terminal, copy default_manager_agent_id from the
creation response into AGENT_ID. Attach to that agent:
plori conversation attach "$AGENT_ID"
That is the same conversation the web app shows, with history and streaming output.
Ctrl-D detaches and leaves the run going. The CLI can approve an action with either
sign-in, except in a run that an MCP client started. To approve those, open the web
address that the CLI prints. The
CLI guide has the full command list, token location, CI
authentication, and JSON output.
To connect from Claude Code, paste this into your conversation:
Set up https://plori.ai/SKILL.md
If Claude asks, type /reload-plugins to load the MCP server, then continue in the
same conversation. Follow the sign-in and approval steps. No plugin installation is
required.
4. Pay only when you outgrow the free allowance
Run time costs $0.01 per minute, billed by the second. Plori charges it to your credit balance. The Plori Router picks the model from a pool shared by every plan. The pool changes as models are released and repriced.
plori warns you before the credit balance reaches zero and links to billing. Stay on Free while it fits. When it does not, compare plans: paid plans add more agents, active workflows, and larger disks. All registered plans include unlimited concurrent agent runs. To use a specific model, including frontier models such as Claude Opus 4.8 and GPT-5.6 Sol, bring your own key. For a temporary spike, add one-time credits without changing plans.
Idle agents sleep and stop billing for compute. Their saved Workspace files and memory remain, so the next task starts from the saved files. Installed software and running processes do not continue.
Composer shortcuts
The message box in the web terminal reads three prefixes while you type. Each one opens a menu that you move through with the arrow keys and accept with Tab or Enter.
/lists the commands this agent accepts.@lists your other agents and inserts the name you pick.$lists the skills this agent has.
A skill is a written procedure the agent reads before it starts work. plori supplies skills for common jobs, such as publishing a site, and the agent writes more as it learns your tasks.
To use a skill:
- Type
$in the message box. - Type part of the skill name or description. The menu shows fewer items.
- Press Tab. The composer writes
$nameinto your message. - Write the rest of your message and send it.
The agent loads that skill for the turn, so you do not describe the procedure again. To close the menu without picking anything, press Escape.
The menu stays closed for a $ inside a word, for a price such as $20, and for a shell
variable such as $PATH. The agent receives these as you wrote them. An agent that has
not run yet has no skills, so the menu is empty until after its first task.