Install The CALL-E Claude Code Plugin

The Claude Code plugin provides the /calle:calle skill. It uses the shared calle CLI so authentication can be checked and recovered from Claude Code without opening the native remote MCP authorization menu.

Prerequisites

Install a Claude Code version that supports plugins and local shell command execution. Check your version with:

claude --version

Install

In Claude Code, add the latest released marketplace from this repository:

/plugin marketplace add https://github.com/CALLE-AI/call-e-integrations.git#@call-e/claude-plugin@latest

Then install the plugin:

/plugin install calle@call-e-claude

Reload plugins in the current Claude Code session so the installed skill is available without restarting:

/reload-plugins

@call-e/claude-plugin@latest is a Git tag updated by the release workflow after @call-e/claude-plugin publishes. For a reproducible install, replace it with a package-level release tag such as @call-e/claude-plugin@<version>.

Authorize

The plugin checks authentication when /calle:calle is invoked. To pre-authorize before using the skill, run:

npx -y @call-e/cli auth login

The command opens the CALL-E browser authorization flow, waits for completion, then stores the token in the private local CLI cache. To verify setup:

npx -y @call-e/cli auth status
npx -y @call-e/cli mcp tools

The plugin uses the repository-local CLI when available, then a global calle, then the npm fallback above. CLI commands run by the skill include this CALL-E attribution:

CALLE_SOURCE=claude CALLE_INTEGRATION=claude_code_plugin CALLE_INTEGRATION_VERSION=0.2.2

Use

Invoke:

/calle:calle

Claude Code will use CALL-E for setup checks, call planning, planned call execution, and call status checks. If authentication is missing or expired, the skill runs blocking calle auth login, shows the browser authorization URL, and continues after authorization completes.

More

See packages/claude-plugin/README.md for package layout and local validation.