Connect Coldvio to your AI agent

Before you start

A Coldvio customer account with access to the workspace you want to use.
Node.js 24 or newer and npm available in your terminal.
A browser on the same machine for the secure sign-in flow.

1. Sign in

The CLI opens Coldvio in your browser. Standard login requests read-only access. Review the exact scopes before approval; use --full-access only when CRM writes, generation, or approvals are needed. Coldvio returns a short-lived, one-time PKCE code to the CLI; your password is never entered in the terminal.

bash

npx -y @coldvio/[email protected] login
npx -y @coldvio/[email protected] status

# Optional after reviewing all four scopes
npx -y @coldvio/[email protected] login --full-access

Release note: these commands become installable when @coldvio/cli is published. Until then, use the package from the Coldvio repository in an approved development environment.

2. Add the MCP server

Choose your client below. The configuration starts a local stdio process and reads your credential from the local Coldvio config directory; never paste a token into a project file.

bash

# Claude Code
claude mcp add --transport stdio --scope user coldvio -- npx -y @coldvio/[email protected] mcp

# Codex CLI and IDE extension
codex mcp add coldvio -- npx -y @coldvio/[email protected] mcp

3. Verify with a read-only request

Ask your agent: “Use Coldvio to show my account summary, then list at most five running campaigns.” The agent should discover get_summary and list_campaigns without needing a token in the prompt.

bash

npx -y @coldvio/[email protected] contacts list --limit 5
npx -y @coldvio/[email protected] campaigns list --status running --limit 5

What to try next

Draft a reply in your voice without sending it.
Summarize pending campaign actions before deciding what to approve.
Find contacts that need follow-up and prepare a review list.
Read the security model before enabling write or approval tools in unattended agents.