Skip to main content
This page is for Claude for Work Owners and Primary Owners (or members with the Identity & Access custom role). If you’re an individual Claude Free, Pro, or Max user, you can add Pleo yourself. See How to Install the Pleo MCP in Claude.

​
Steps

Follow the steps below in the Claude Admin Console to add Pleo as a custom connector for your organisation.
  1. Go to Organization settings → Connectors.
  2. Click Add, then choose Custom → Web.
  3. Enter the Pleo MCP server URL:
No Advanced or Auth settings are needed.
  1. Add the connector. It becomes available to everyone in your organisation.
  2. Set Tool permissions for Pleo’s tools: Always allow, Needs approval, or Blocked. Pleo recommends requiring approval for any tool that writes data.
Pleo is now available as a connector for your organisation. See Anthropic’s documentation: Add a custom connector

​
Allow Network Access for Receipts

The agent reaches two Pleo domains directly from the commands it runs:
  • Uploading a receipt: Pleo issues a short-lived upload URL on mcp.pleo.io and the agent uploads the file to it.
  • Reading a PDF receipt: Pleo returns a short-lived link on receipts.cdn.pleo.io and the agent downloads the file from it. Image receipts come back inside the tool result and don’t need this domain.
If the command’s network access doesn’t allow these domains, the upload is blocked and the agent can’t read PDF receipts. For how this works across AI clients, see Allow Network Access for Receipts.

​
Claude Code

Claude Code’s sandbox is off by default, and without it no change is needed. If you use the sandbox, it pre-allows no network domains for the shell commands it runs, so add mcp.pleo.io and receipts.cdn.pleo.io to its allowed domains:
Keep any domains you already allow and add these to the list, rather than replacing it. This setting covers the commands the agent runs, which is what uploading and reading receipts use. Claude Code’s connection to the Pleo MCP is made by the app itself and doesn’t depend on it. *.pleo.io works if you prefer a broader rule, though these two domains are all the agent needs. If the sandbox is on but the Pleo domains aren’t listed:
  • By default: Claude Code asks for approval the first time a command needs the domain, and the developer can allow it from that prompt.
  • In auto mode: Claude lists the hosts on the command itself and the auto mode classifier reviews them with the command, so no network prompt appears.
  • With strictAllowlist: no prompt appears and an unlisted domain fails outright, so the entry is required.
  • With allowManagedDomainsOnly: only managed entries count. Add mcp.pleo.io and receipts.cdn.pleo.io to the managed allowedDomains, because an entry in ~/.claude/settings.json is ignored.
See Anthropic’s documentation: Configure the sandboxed Bash tool

​
Claude Web, Desktop, and Cowork

When Claude runs code, for example to upload a receipt file you’ve shared in the conversation or to read a PDF receipt, that code only reaches the domains your network egress setting allows. The connection to the Pleo MCP itself isn’t affected by this setting. On Team and Enterprise plans, an Owner changes this in Organization settings → Capabilities:
  1. Make sure Code execution and file creation is on.
  2. Turn Allow network egress on. It’s off by default for new Enterprise organisations, and Team organisations start with package managers only.
  3. Choose Allow network egress to package managers and specific domains and add mcp.pleo.io and receipts.cdn.pleo.io.
If egress is already set to All domains, no change is needed. On Free, Pro, and Max plans, you manage code execution and network egress yourself in Settings → Capabilities. Cowork sessions in the cloud can also have their own network-access policy in organisation settings. See Anthropic’s documentation: Create and edit files with Claude and Claude Cowork architecture overview

​
What Comes Next?