up to $20
For developers · MCP server live

Let your agent delegate.

Connect Claude Code, Codex, Cursor or any MCP client to Plori. Your agent can create cloud Workspaces, send them tasks and read the replies. You sign in through the browser; there is no API key to manage.

  • Several agents in one Workspace. Your agent can start several Plori agents on the same project at once. Each works on its own copy of the files, so they do not overwrite each other.
  • Plori merges changes and handles conflicts.
  • Another agent is quick to start. It begins from the Workspace's current files, so your client uploads and clones nothing.
What it can do

Set up https://plori.ai/SKILL.md

Connect and run your first task.

§01

One endpoint: https://api.plori.ai/mcp (Streamable HTTP). Choose your client. Clients that support OAuth sign you in through the browser. Scripts and CI use an API key, described under Other ways to connect.

  1. 1

    Paste this into Claude Code. It guides you through sign-in.

    Set up https://plori.ai/SKILL.md
  2. 2
    Claude shows a short code to enter at plori.ai/device on your phone or computer, or it opens a browser sign-in. If Claude asks, type /reload-plugins in the conversation.

No plugin to install.

Try your first task

After you connect, paste this into your client.

“Ask Plori to research ways to reduce household food waste and save a one-page summary in a Workspace.”

Other ways to connect.

§02

plori CLI (terminal)

Not an MCP client. Sign in through the browser, then attach to a live agent session in your terminal.

curl -fsSL https://plori.ai/install.sh | sh
plori login && plori attach <agent-name>

CI, scripts, and headless use

Create a key in Settings → API keys and send it as Authorization: Bearer.

{
  "mcpServers": {
    "plori": {
      "type": "streamable-http",
      "url": "https://api.plori.ai/mcp",
      "headers": { "Authorization": "Bearer plori_sk_…" }
    }
  }
}

What your agent can do.

§03
  • Create Workspaces and send them tasks. The Workspace's agent replies in its conversation.
  • Send one task and let the Workspace's agent split the work across several agents and merge the results.
  • Start several Plori agents itself. Plori merges their changes.
  • Read the Workspace's files and its saved versions.
  • Follow a run, cancel it, or answer a question the agent asks.
  • Schedule a run for later.
  • Check the credit balance and what a Workspace's runs cost.

Every call acts only within the account that signed in. Workflows, disk and usage details stay in the web app, the REST API and the CLI. Tool names and arguments for agents are in /agents.md.

Only files under /workspace are kept between runs. Software a run installs elsewhere is not kept.

Plori sends task messages, the files the agent reads and the agent's work through OpenRouter to OpenAI, DeepSeek, Anthropic and Moonshot AI, which supply the agent's models and its reviews. When the agent generates an image, the prompt and any reference images go to an image model provider through OpenRouter. TypeSafe receives typed-judgment requests about the agent's steps, which can include images from the Workspace. Runs that start through the MCP server do not generate video. Video generation is available in the web app and the CLI. The provider list follows Plori's model catalog and can change.

What it costs, what it can spend.

§04

Free to connect, credits to run

Connecting and reading are free. Running an agent spends credits. Run time costs $0.01 per minute, billed by the second while an agent is awake, on every plan. Idle agents sleep and stop billing for compute.

Model usage draws credits at cost with no markup, based on the provider's real price. If a platform fault stops a run, Plori refunds its charges, subject to the per-account daily limit. Bring your own model key and model usage drops to $0. Rates and worked examples are on the pricing page.

Scoped to your account

Every call acts as the account that signed in, and nothing else. Concurrent agent runs are unlimited on every plan. Cut access any time: delete the API key in Settings, or remove the server from your client. OAuth sign-ins expire on their own.

What you confirm yourself.

§05

A run started through MCP, or by a schedule an AI client created, pauses before any action outside the Workspace: a write through a connected account or a connected MCP tool, an outbound HTTP write, a git push, a package publish, a deploy, an upload, an email, a public tunnel, an irreversible browser action, or activating or running a workflow that writes. Reads do not pause.

You approve each action, one at a time, in the Plori web app. Your agent cannot approve it: it can only deny the request or answer a question. “Always allow” grants do not apply to these runs. A scheduled run is queued only after you confirm it in the web app, and only you can make a site public.

In the browser, too.

§06

The homepage and the signed-in web app expose WebMCP tools, so a compatible agentic browser can check credits, open an agent, send a task and follow the run.

Tell us what you're building.

§07

Using plori over MCP, or missing something? Send a note to the team.

to send feedback.

Prefer to talk? Join our Discord or email [email protected]. Building against the API directly? Start at /agents.md.