← All integrations

AI assistant integration

Connect Pi
to Network Perspective

Ask Pi to explore your team’s reports and metrics through your own Network Perspective connection.

Before you start

  • Pi with built-in remote MCP support. These settings were checked against Pi 1.4.0. Run the browser and CLI on the same machine.
  • A Network Perspective account with access to the company you want to explore. Ask your Network Perspective administrator to enable AI assistant access if it is unavailable.
  • Your organization’s approval to use this tool with your data and allow its connection to Network Perspective.
Your permissions stay with you.

The connection reads the reports and metrics available to your account, including report-visible comments and quotations. It does not change reports or settings.

1. Add Network Perspective

Add the user configuration

Create or edit ~/.pi/agent/mcp.json to make the connection available to your Pi sessions.

If this file already exists, merge the np entry into its existing configuration. Preserve your other servers and settings. These values are public setup identifiers; no client secret is needed.

ConfigurationCopy
{ "mcpServers": { "np": { "url": "https://app.networkperspective.io/mcp", "oauth": { "clientId": "np-pi", "callbackUrl": "http://127.0.0.1/callback", "scope": "np.read offline_access" } } } }

Choose the configuration scope

For a project-only setup, use .pi/mcp.json in the project root and approve project trust. A project entry named np replaces a user entry with the same name. Use one location intentionally; do not maintain conflicting copies.

2. Connect your own account

Run this command in your terminal:

Sign inCopy
pi mcp login np

If Pi is already running, use /reload after editing its configuration. Use /mcp to check the connection.

  1. Sign in with your own Network Perspective account in the browser.
  2. Choose the intended company or network when offered, review the requested access, and approve the connection.
  3. Return to your tool and check that Network Perspective is connected.

Each connection is authorized for one company. To work with another company, use a separate connection where your tool supports it.

3. Try the connection

Start a conversation in Pi and ask:

First questionCopy
Use Network Perspective to tell me which company I’m connected to and the current reporting month.

Check the answer: the assistant should use Network Perspective’s get_company_info tool and return the company you intended to connect and its current reporting month. A server appearing in a list alone does not confirm it is working.

Then try: “What metrics are available?”

If something gets in the way

Pi does not find the server

Check the configuration location and JSON syntax. A project-level np entry can replace your user settings. Reload the session after changes.

A callback or scope error appears

Use scope, singular, and the callback shown above. Pi supplies a free local port; no fixed port is needed. If you use another Pi version, check its MCP configuration format before changing these values.

Authentication expires

Run pi mcp login np again with your own account, then check the connection in /mcp.

Sign-in or company access fails

Check the exact client ID in this guide. If the client is rejected, contact Network Perspective. For a missing company or access denied, ask your Network Perspective administrator to check membership, AI assistant access, and report permissions.

The callback does not match

Contact support with the client version and callback value only. Do not send a full authorization URL, sign-in code, token, or secret. Do not change callback hosts or paths to work around an error.

Disconnect whenever you need to

In Network Perspective, open Me → Connected agents, find the connection, and disconnect it. This revokes that connection’s access.

Removing it from your AI tool or clearing local files alone does not revoke access in Network Perspective. Disconnecting cannot remove information already shared with the AI provider.

Open Connected agents →