The official Claude Code marketplace for TruVerifAI.
Submit each slash command on its own — Claude Code parses one slash command per submission. Pasting them all together produces a malformed-URL error.
1. Add the marketplace:
/plugin marketplace add https://github.com/TruVerifAI/claude-plugins.git
2. Install the plugin:
/plugin install truverifai@truverifai
3. Reload Claude Code's plugin set:
/reload-plugins
4. Configure your API key. Run /plugin, click Installed →
TruVerifAI, paste your tvai_… key (generate one at
truverif.ai/settings/api-keys),
click Save configuration, then run /reload-plugins again.
5. Enable auto-update (recommended). While you're in the /plugin
UI, click the Marketplaces tab, select truverifai, and toggle
Enable auto-update on. This is a one-time setting — once enabled,
future plugin updates (new skills, bug fixes, new tools) flow in
automatically on the next Claude Code session start. Without this,
you'd have to run the three-command manual update sequence below
every time a new version ships.
6. Verify everything's wired up: run /truverifai-setup.
If you didn't enable auto-update, pull the latest release with these three commands (submit each on its own):
/plugin marketplace update truverifai
/plugin install truverifai@truverifai
/reload-plugins
To remove the plugin (and optionally the marketplace registration), submit each slash command on its own:
/plugin uninstall truverifai@truverifai
/reload-plugins
Optional — remove the marketplace registration too:
/plugin marketplace remove truverifai
| Plugin | Description |
|---|---|
truverifai |
Multi-model second-opinion deliberation for high-stakes coding decisions. Five skills — three primary (audit / deliberate / synthesize) that route your agent to TruVerifAI's MCP server, a follow-up (record-outcome) that fires after acting, and a gate-release skill (skip-gate-when-not-needed) — plus proactive PreToolUse review gates that prompt an audit before risky commits and a deliberation/synthesize before risky design changes (cross-domain risk classifier; releasable by a review or a logged record_gate_skip). Four frontier models reason independently and conflict-target each other's positions to produce decision-grade output; outcome reporting closes the loop with measurable per-call impact data on the dashboard. |
For a walkthrough of how skills + adherence telemetry work after install: truverif.ai/settings/mcp (sign in first).
For non-Claude-Code clients (Codex CLI, Goose, Cursor, Claude Desktop, VS Code, Cline), the same page has manual MCP server setup snippets and trigger-rule paste targets.
- Email: support@truverif.ai
- Issues: github.com/TruVerifAI/claude-plugins/issues
MIT — see LICENSE inside each plugin.