As you know, we have been hard at work over the past several months on the Agent Metadata Specification. Thanks to many of you for your contributions to both the Cross-Industry and Industry versions of the spec.
Agent Metadata Specification includes Business Context for the Agent
Here is a simplified conceptual model of the latest version of the spec with a focus on Business Context that ties the Agent to the AI Use Case, Application, AI Model, Business Process, Physical AI, and Provider (see image above).
Answering What-If Questions
Once we tie the Agent Business Context with the related Applications, Devices, Regulations and Processes, you are able to answer “what-if” questions in minutes.
Utilities Example: Leveraging Agent Metadata to Produce a Business Continuity Plan in Minutes
Here is an example where Tavro combined our Open-Source Agent Catalog based on the Agent Metadata Specification and an MCP Server to produce the Business Continuity Plan (BCP):
- Conceptualized the Transformer Prediction Agent
We prototyped the Transformer Prediction Agent that we have discussed earlier in the Utilities industry. This agent maps to all the relevant artifacts including AI Use Case, Regulation, Tools, Data Source, Risk Assessment, Application, and Business Process.
Prototyped the Agent in Tavro’s Open-Source Agent Catalog BEFORE Development
We then prototyped the agent in Tavro’s open-source agent catalog before it was even developed. This reduces unnecessary back-and-forth and substantially reduces the ideation process.Here is the Agent Context Diagram in Tavro.
We also prototyped agent instructions in Tavro. This makes it easier to conduct an initial agent risk assessment as we will see later.
- Conducted Initial Agent Risk Assessment Using Tavro Agents
We used Tavro’s own AI Risk Assessment Agents that classified the agent as High Risk based on Article 6 of the EU AI Act (Critical Infrastructure).
Tavro’s Agents also developed an Agent Risk Assessment based on OWASP AIVSS.
- Used Tavro’s MCP Server with Claude to Generate an Agent BCP
We used Tavro’s MCP Server with Claude to generate an Agent BCP in minutes (we tested OpenAI too). The BCP was based on a detailed understanding of the associated applications like SAP Plant Maintenance, data sources,and tools. A Human-in-the-Loop still needs to update the report but this likely saved weeks of effort.
The BCP also identified detailed gaps requiring resolution.
Here is a link to the entire BCP in the Agent Metadata Specification GitHub repo.
Links to Tavro Open-Source Agent Catalog & Free Trial
Here is a link to Tavro’s Open-Source Agent Catalog. You can also request a free trial hosted by Tavro.