Identity and Runtime Authorization for AI Agents

Highflame is a runtime control plane that sits between AI agents and the models, tools and data they reach.

It discovers agents across clouds, developer IDEs and SaaS tools, assigns each a verifiable identity, and authorizes every LLM call, MCP request and tool action against policy, escalating sensitive actions for human approval.

Decisions draw on more than 150 runtime signals and are logged as signed, tamper evident records mapped to OWASP, NIST, MITRE and the EU AI Act.

The platform is built for security teams that must show who each agent acts for and what it is allowed to do.

Based in San Francisco and led by CEO Sharath Rajasekar, Highflame released its identity layer, ZeroID, as open source in April 2026.