IronVise

See IronVise in action

Watch how AI actions become provable audit evidence

IronVise records every AI decision, blocks unsafe actions before execution, and produces tamper-evident certificates you can verify offline.

Before you start

What this demo shows in under 10 minutes

  • How agent actions are captured in a tamper-evident timeline.
  • How the Claude Code hook and Tripwire can block unsafe operations before execution - and that the recording proxy only observes.
  • How evidence is exported for independent verification.
Security Leads Compliance Teams Engineering Managers Auditors

Demo mode is optimized for fast evaluation. For production use, configure guard, policy, backend, and webhook settings explicitly.

Before you run demo

Prerequisites

What you need depends on how you run the demo. At minimum, choose one path below.

  • Rust 1.94+ - All native launcher and CLI demos (rustup recommended).
  • Docker - Only for the Docker demo (docker-compose.demo.yml). Not required for make demo, SQLite, or embedded PostgreSQL.
  • LLM backend - At least one: Ollama (local), OpenAI, or Anthropic. For audit and orchestrate the primary LLM must be reachable; the Guard uses GUARD_LLM_BACKEND / GUARD_LLM_MODEL when GUARD_REQUIRED=true.
  • Node.js 20.19+ and npm - Only if you build or run the Tauri 2 desktop GUI (gui/). The Docker demo exposes the REST API and Observer only; no GUI.

How to use demo mode

Different ways to run the demo

You can try IronVise with zero config (embedded Postgres, auto LLM detection) or use Docker for a fully containerized stack.

Demo defaults prioritize fast evaluation. Before production rollout, explicitly configure guard, policy, backend mode, and webhook settings.

1. make demo (zero config)

Starts the Observer dashboard against a local SQLite ledger with ECTO_DEMO_MODE=1. No Docker, no PostgreSQL, and no API key required - the ledger file is created on first run.

# macOS / Linux
make demo

# Windows (PowerShell)
.\ironvise-win.ps1 -demo

Equivalent to ECTO_DEMO_MODE=1 DATABASE_URL=sqlite://ectoledger-demo.db cargo run --release -p ectoledger -- serve. The ledger lives in ectoledger-demo.db in the repository root; delete that file to wipe demo data. Demo mode is insecure by design - see running securely before any real deployment.

2. make start (desktop GUI)

Runs the Tauri desktop app, which starts its own Observer backend in-process - no separate server to manage. Uses your configured DATABASE_URL and LLM. Good for development and real audits. Use make backend for the server alone.

make start          # macOS / Linux - Windows: .\ironvise-win.ps1

3. Docker demo (no Rust or Node)

One command pulls the pre-built image and starts PostgreSQL 17, Ollama, and IronVise. Open http://localhost:3000 for the Observer dashboard. No Tauri GUI in Docker - use make start for the full desktop experience.

docker compose -f docker-compose.demo.yml up

To reset: docker compose -f docker-compose.demo.yml down -v

4. SQLite + serve (no Docker)

For a local dashboard or CLI audit with no external database, set DATABASE_URL=sqlite://ledger.db and run cargo run -- serve (or audit, report, verify-session, anchor-session). Only multi-agent orchestrate, diff-audit, and red-team require PostgreSQL.

Full details: Zero-Config Demo and Demo Readme on GitHub.

Troubleshooting

Common issues and fast fixes

Audit command fails

  • Confirm backend mode supports your command.
  • Check primary LLM connectivity and API keys.
  • Verify guard settings when running production mode.

Demo differs from production

  • Demo uses isolated defaults for quick setup.
  • Set explicit production env vars before deployment.
  • Validate policy, tripwire, and webhook config in production.

Verification confusion

  • Chain verification checks event integrity.
  • verify-cert validates certificate evidence offline.
  • VC endpoint checks differ from full signature verification.
IronVise GUI demo

Feature Clips

What IronVise actually does

IronVise demo gif

Explore an audit trail

Walk through recorded actions, decisions, and verification using demonstration data.

Audit evidence in the current interface

Inspect example policy decisions, data receipts, and session verification in IronVise 0.8.

Interface walkthrough

Dashboard screens and control surfaces

Dashboard screen

Give your AI agent a task and watch every decision get recorded.

Audit history screen

Review exactly what your AI did, step by step.

Tripwire screen

Define boundaries your AI cannot cross.

Policies screen

Write rules that govern what your AI is allowed to do.

Prometheus metrics screen

Monitor AI agent behavior and reliability in real time.

Live feed screen

Watch the pre-execution checks evaluate each action as it happens.

Next step

Ready to run your first audit?

Start with Quick Start for setup, or review the latest release notes before integrating IronVise.