Agentnotary keeps your artificial intelligence agents safe and compliant. The software helps you manage, seal, and verify the actions performed by your AI systems. It provides a formal record for every decision made by the agent. This process helps you meet legal standards like the EU AI Act. It also tracks the components inside your agents so you know exactly what tools they use.
- Cryptographic Sealing: Every agent process receives a digital seal to confirm its identity and integrity.
- Runtime Protection: Guards monitor the agent in real time to stop unauthorized actions or unexpected behavior.
- Compliance Reports: The app creates documentation to show that your AI follows safety and ethical standards.
- AI-BOM Creation: It generates a list of every library and component used, similar to a digital manifest.
- Adversarial Testing: Built-in tools check your agents for weaknesses against common attack tactics.
You must follow these steps to set up the software on your Windows computer.
- Check Requirements: Ensure your system runs Windows 10 or Windows 11. You need at least 4GB of RAM and an active internet connection.
- Visit the Download Page: Go to this link to find the latest version of the installer.
- Download the File: Find the file ending in
.exeunder the "Assets" section of the latest release. Click the file name to start the download. - Launch the Installer: Open your Downloads folder. Locate the file you just downloaded and double-click it.
- Follow Prompts: Your computer might show a security warning. This happens because the installer is new. Click "More info" and then select "Run anyway" if you trust the source.
- Complete Setup: Follow the instructions on the screen to finish the installation.
The application interface allows you to connect your AI agent configuration files. Once the software reads your configuration, it scans the agent for vulnerabilities. You can press the "Seal" button to register the agent's current state. This creates a snapshot that proves the agent has not changed since the last audit. If the agent acts outside its provided rules during runtime, the tool flags the event for your review.
Artificial intelligence behaves in complex ways. Sometimes, these systems make mistakes or follow bad instructions. Agentnotary acts as a supervisor. It ensures the inputs and outputs of your agents stay within safety limits.
The software uses a database of known threats to check your agents. This protects you from issues like prompt injection, where an outside user tries to trick the agent into doing something it should not. The tool maps your agent's behavior against industry standards to ensure you meet safety goals.
Even if you do not understand the underlying code, the software manages these technical systems for you:
- CycloneDX Integration: The app uses this standard to track your software components. It ensures your AI-BOM is easy to read.
- SPDX Support: You can export your data in standard formats for external audits.
- Guardrails: The software enforces rules during the interaction between your agent and the Large Language Models.
- CLI Access: Advanced users can perform these same actions using commands if they prefer to use the command prompt.
If the application fails to start, verify that your user account has administrative privileges. Sometimes, antivirus software blocks new applications. You may need to add an exception for the tool in your antivirus settings.
If the agent scan takes a long time, check your internet connection. The tool may need to fetch the latest threat intelligence data from secure servers. If the scan still hangs, restart the application and try again.
Does this tool work with private agents? Yes. All processing happens on your local machine. The tool does not send your private data to external servers.
What happens when the agent fails an audit? The software will produce a red alert in the dashboard. You can click this alert to read a report on which rule the agent broke.
Does this software cost money? The core functionality remains available to all users at no cost. You can view the source code on GitHub to verify how the application works.
Do I need a persistent internet connection? A connection is required only when you want to update the tool or download new safety definitions. The agent auditing works perfectly while offline.
The European Union sets strict rules for AI transparency. Agentnotary simplifies this by automating the documentation process. You do not need to write technical manuals manually. The app gathers the information, formats it into a report, and stores the history for your records. This creates a verifiable trail for regulators and business stakeholders.
Check the release page periodically for new versions. Each release contains improvements for security and performance. Updating is simple: just download the new installer and run it over the old version. Your current settings and agent history will remain intact.
Always keep your agent API keys secret. Do not share your keys with anyone. This application helps secure your agent's behavior, but it cannot fix issues if someone steals your access credentials. Store your keys in a secure environment.
After setting up the tool, keep track of your audit logs. Regularly review the "Audit Summary" tab to see if your agents are using secure libraries. If a new version of a library comes out, use the "Refresh BOM" button to update your documentation. This ensures your compliance records remain current at all times.