Official AgentGuards plugin marketplace — LLM security guardrails for AI coding agents: jailbreak and prompt-injection detection, web-content scanning, data-exfiltration blocking, and destructive-command authorization. Enforcement is configurable — fail-closed by default, or fail-open (availability-first) with a single environment variable.
| Plugin | Agent | Deployment | Version | Description |
|---|---|---|---|---|
agentguards-claude |
Claude Code | hosted | 0.2.29 |
MCP server + enforcing hooks (input, Bash, web-content) and security instructions. |
agentguards-codex |
OpenAI Codex | hosted | 0.2.13 |
Enforcing hooks (input, shell, web-content) + MCP server and security instructions. |
agentguards-gemini |
Gemini CLI | hosted | 0.1.8 |
MCP server + enforcing hooks (input, tool-call, web-content) and security instructions. |
agentguards-copilot |
GitHub Copilot CLI | hosted | 0.1.6 |
MCP server + enforcing hooks (input, shell, web-content) and security instructions. |
@agentguardsco/opencode-plugin |
OpenCode | hosted | 0.1.6 |
Enforcing plugin (prompt, bash, web-content), + MCP server and security instructions. |
agentguards-claude-selfhosted |
Claude Code | self-hosted | 0.1.11 |
Hooks only — no bundled MCP server, and no default URL, so it can never talk to the hosted service by accident. |
agentguards-codex-selfhosted |
OpenAI Codex | self-hosted | 0.1.6 |
Hooks only — no bundled MCP server, and no default URL. |
agentguards-gemini-selfhosted |
Gemini CLI | self-hosted | 0.1.5 |
Hooks only — no bundled MCP server, and no default URL. |
agentguards-copilot-selfhosted |
GitHub Copilot CLI | self-hosted | 0.1.4 |
Hooks only — no bundled MCP server, and no default URL. |
Versions above are the current release in this repo; .github/workflows/plugin-integrity.yml fails the build if they drift from the plugin manifests. To see what you actually have installed, use your agent's own listing (/plugin in Claude Code).
/plugin marketplace add alelaguard/agentguards-plugins
/plugin install agentguards-claude@agentguards
Then set your API key (get one at https://agentguards.co/dashboard/keys):
export AGENTGUARDS_API_KEY=ag_your_token_here
Add that to your shell profile and restart Claude Code, or run
/agentguards:setup. See claude/README.md for full
configuration.
Alternative: install via npm. The claude/ plugin is also published as
@agentguardsco/claude-plugin
for programmatic use — pinning an exact version in package.json, CI
provisioning, or embedding the hook script in your own tooling — outside of
Claude Code's interactive /plugin flow:
npm install @agentguardsco/claude-plugin
Note this only fetches the plugin's files; it does not register hooks,
skills, or the MCP server with Claude Code (that wiring happens through
/plugin install above). Use the npm package when you need the raw files,
use /plugin install when you want it running in Claude Code.
codex plugin marketplace add alelaguard/agentguards-plugins
Enable the agentguards-codex plugin, then set your API key (get one at
https://agentguards.co/dashboard/keys):
export AGENTGUARDS_API_KEY=ag_your_token_here
Add that to your shell profile and restart Codex. See
codex/README.md for full configuration.
Gemini CLI's extensions install 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 API key (get one at https://agentguards.co/dashboard/keys):
export AGENTGUARDS_API_KEY=ag_your_token_here
Add that to your shell profile and restart Gemini CLI. See
gemini/README.md for full configuration.
copilot plugin install alelaguard/agentguards-plugins:copilot
Then set your API key (get one at https://agentguards.co/dashboard/keys):
export AGENTGUARDS_API_KEY=ag_your_token_here
Add that to your shell profile and restart Copilot CLI. See
copilot/README.md for full configuration.
opencode plugin @agentguardsco/opencode-plugin
Then set your API key (get one at https://agentguards.co/dashboard/keys):
export AGENTGUARDS_API_KEY=ag_your_token_here
Add that to your shell profile and restart OpenCode. See
opencode/README.md for full configuration, including
MCP server setup (a separate step for OpenCode, unlike the other agents above).
MIT
