Updated
Agent Plugins
LyraShield ships as a portable plugin following the open Agent Plugins v1.0.0 standard. One package bundles two things: the LyraShield MCP server, which gives your agent the scan, finding, fix-plan, and verify tools, and a security skill that tells the agent when to reach for them. For the five launch clients below, this replaces hand-editing an MCP config file.
Install in two commands
Log in first. The bundled MCP server runs over stdio and reads your selected workspace credential from ~/.lyrashield/credentials.json, so the login has to come before the install or the plugin has no credential to read.
npx lyrashield login --oauth
npx lyrashield install <agent>Prefer to let the CLI decide? npx lyrashield init detects the agents you have installed and picks the plugin for the clients that support it, falling back to the verified config-file or vendor-CLI path for the rest. Run npx lyrashield doctor afterwards to confirm what is configured, and add --dry-run to any install to preview the change without writing.
~/.lyrashield/credentials.json, which is user-only. A workspace API key in LYRASHIELD_API_KEY remains the fallback for CI and clients that cannot complete OAuth. Do not commit either credential.The five launch clients
An Agent Plugin exists for these five clients only. Every other agent LyraShield supports uses the config-file, vendor-CLI, or guided-manual path in its own guide, and those guides stay the reference for exactly what gets written.
| Client | Plugin location | CLI default |
|---|---|---|
| Claude Code | ~/.claude/plugins/lyrashield | Agent Plugin |
| Cursor | ~/.cursor/plugins/local/lyrashield | Agent Plugin |
| OpenAI Codex | ~/.codex/plugins/lyrashield | Agent Plugin |
| GitHub Copilot | ~/.copilot/plugins/lyrashield | Agent Plugin |
| Kiro | ~/.kiro/plugins/lyrashield | Agent Plugin |
npx lyrashield install vscode or follow the VS Code guide.What the skill adds
The MCP server alone gives an agent tools; the bundled skill gives it judgement about depth. It maps a request like a pre-commit diff check to a quick bounded review, and a deep security review to the intrusive agentic mode that only runs inside the authorised isolated sandbox for a repository target. That keeps a routine check from silently becoming a long expensive scan.
Findings are proposals, not automatic changes
Installing the plugin does not let an agent rewrite your code unattended. Read-only tools work with any valid credential. Mutating tools need a write-scoped credential, and over the remote endpoint they also need a human out-of-band approval: the server returns a pending result with an approval link rather than executing, and the agent can only proceed after a person approves it. See Approvals for the full flow. Fixes LyraShield generates are proposals requiring review before they land.
Uninstall
Removing the plugin is the reverse of installing it. The CLI deletes the plugin directory it created and leaves the rest of your agent's configuration untouched.
npx lyrashield uninstall <agent>Frequently asked questions
What is the LyraShield Agent Plugin?
It is a portable plugin that follows the open Agent Plugins v1.0.0 standard published at agent-plugins.org. A single package carries both the LyraShield MCP server and a security skill that tells the agent when to run LyraShield checks, so a supported client gets the tools and the guidance from one install instead of a hand-edited MCP config.
Which clients can install the Agent Plugin?
Five launch clients: Claude Code, Cursor, OpenAI Codex, GitHub Copilot, and Kiro. Every other supported agent uses the verified config-file, vendor-CLI, or guided-manual path documented in its own guide. VS Code and Windsurf use config-file installation.
Do I need to log in before installing the plugin?
Yes. Run npx lyrashield login --oauth first. The bundled MCP server runs over stdio and reads the selected workspace credential from ~/.lyrashield/credentials.json, so installing before you log in leaves the plugin with no credential to use. The CLI reports this rather than writing a broken install.
Does the plugin let the agent change my code on its own?
No. LyraShield findings and fixes are proposals requiring review. Mutating tools need a write-scoped credential, and over the remote endpoint they additionally need a human out-of-band approval before they run.
Why does VS Code use an MCP config file instead of the plugin?
No generated VS Code Agent Plugin shim exists. Its verified path is the .vscode/mcp.json file that npx lyrashield install vscode writes. The VS Code guide documents the servers root key and per-entry stdio type.