Integrations
The plugin is the best and fastest way to protect each coding agent — one install adds the enforcing hooks and the guardrail skills (the Gemini CLI and Copilot CLI plugins also include the MCP server). Grab an API token from your dashboard, then pick your agent below.
Prefer to watch it set up? See setup demos on our YouTube channel →
Which one do I need?
Install one path per tool. Automatic means every prompt and command is checked; on request means the AI calls AgentGuards when it decides to, so a check can be skipped. Adding a second path to the same tool checks everything twice and uses your quota twice.
| You use | Install | Checks | Don't also add |
|---|---|---|---|
| Claude Code (terminal, or the Desktop app's Code tab) | Plugin | Automatic | the Anthropic API proxy or the MCP server for the same Claude Code |
| OpenAI Codex CLI | Plugin | Automatic | the AgentGuards app (ChatGPT/Codex directory) in Codex |
| ChatGPT | AgentGuards app | On request | nothing to add: ChatGPT has no hooks, so on request is the most it supports |
| Gemini CLI, GitHub Copilot CLI, OpenCode | Plugin | Automatic | a separate MCP server entry (the Gemini and Copilot plugins already include it) |
| VS Code Copilot, Cursor, other MCP clients | MCP server | On request | — |
| Your own app | REST API / Gateway | Every call you make | — |
Claude Code
# Inside a Claude Code session:
/plugin marketplace add alelaguard/agentguards-plugins
/plugin install agentguards-claude@agentguards
# Then set your token (add to ~/.bashrc or ~/.zshrc) and restart Claude Code:
export AGENTGUARDS_API_KEY=ag_your_token_hereThen run /agentguards:status to confirm it's active. Enforcement is fail-closed by default; set AGENTGUARDS_FAIL_OPEN=true to fail open.
For manual hook setup and full configuration see the Claude Code integration guide →
OpenAI Codex
codex plugin marketplace add alelaguard/agentguards-plugins
codex plugin add agentguards-codex@agentguards-codex
# Then set your token (add to ~/.bashrc or ~/.zshrc) and restart Codex:
export AGENTGUARDS_API_KEY=ag_your_token_hereFor manual hook setup and full configuration see the OpenAI Codex integration guide →
Gemini CLI
Gemini CLI's extension installer only supports single-extension repos, so install by cloning and linking the subdirectory:
git clone https://github.com/alelaguard/agentguards-plugins.git
gemini extensions link agentguards-plugins/gemini
# Then set your token (add to ~/.bashrc or ~/.zshrc) and restart Gemini CLI:
export AGENTGUARDS_API_KEY=ag_your_token_hereFor manual hook setup and full configuration see the Gemini CLI integration guide →
GitHub Copilot CLI
Copilot CLI supports installing straight from a repo subdirectory, so no separate clone step is needed:
copilot plugin install alelaguard/agentguards-plugins:copilot
# Then set your token (add to ~/.bashrc or ~/.zshrc) and restart Copilot CLI:
export AGENTGUARDS_API_KEY=ag_your_token_hereFor full configuration see the agentguards-copilot plugin README →
OpenCode
opencode plugin @agentguardsco/opencode-plugin
# Then set your token (add to ~/.bashrc or ~/.zshrc) and restart OpenCode:
export AGENTGUARDS_API_KEY=ag_your_token_hereFor manual/local install and MCP setup see the OpenCode integration guide →
ChatGPT
ChatGPT has no hooks, so it gets the AgentGuards app instead: ChatGPT calls it to scan shell commands, code and untrusted text when it decides something needs checking. In ChatGPT open Plugins → Add → Add custom MCP server, use this server URL with OAuth, then sign in to AgentGuards and click Connect. No key to paste.
https://prod.agentguards.co/directory/mcpUsing Codex? Install the Codex plugin above instead — adding this app to Codex as well checks everything twice. Step-by-step: ChatGPT integration guide →
Other integration options — VS Code (GitHub Copilot), GitHub Copilot Enterprise, the REST API, and the Gateway API — are available from your dashboard →