IndyKite Open Sources Intent Agent for AI Agent Governance

 IndyKite announced the open-source release of the IndyKite Intent Agent, which translates a person’s natural-language request into a structured intent that every agent and tool in a workflow can be held to. Built on OpenID AuthZEN semantics, the Intent Agent gives developers a standards-based way to express what an AI agent was actually asked to do.

Most AI agents today act on behalf of a person through delegation. That gives each agent its own identity and records who it acts for. On its own, however, delegation hands the agent the full breadth of that person’s permissions, and passes them down every hop of a multi-agent workflow. A request to view an account balance can, three agents later, still reach a payment tool.

Read more –VIACHAIN and Star Solutions Formalize Alliance to Keep Containers Connected at Sea

Intent-based access control addresses this by limiting an agent to the task the person asked for. But intent expressed in natural language cannot be enforced, and approaches that ask an AI model to judge every action produce decisions that are probabilistic by design.

The IndyKite Intent Agent takes a different approach. AI is used once, at the start of a workflow, to translate the request. From that point on, every decision is made by rules.

From plain language to AuthZEN. Each request becomes a structured intent: a subject who wants to perform an action, or set of actions, on specific resources in a given context. “Can I see the refund docs?” becomes subject: this user; action: read; resource: documents of type refund.

Grounded in enterprise data. Intent is extracted against the organization’s own knowledge graph schema and access policies. Requests that fall outside the domain an agent was built for are rejected rather than forced into a match.

Bound to a verified identity. The subject of the intent is taken from the caller’s verified identity, never from the prompt, protecting against attempts to impersonate another user.

Read more – Traefik Labs Expands Partner Ecosystem for AI Governance

Auditable by default. Every generated intent is written to an audit log, so decisions can be reproduced and proven after the fact.

Built on open standards. The Intent Agent uses OpenID AuthZEN semantics, communicates over the Agent2Agent (A2A) protocol, and ships as a container image that runs alongside existing infrastructure.

The IndyKite Intent Agent is available now under an open-source license at github.com/indykite/intent-agent, where developers can contribute, open issues and join discussions. Intent chaining, the Agent Gateway and the Intent-based Access Control engine are available in early-access with IndyKite’s AgentControl. The Intent Agent is being released today at the AI builders Meetup in London.

Read more – Broward County Selects ServiceNow to Deploy AI Agents

About IndyKite
IndyKite is the first System of Intelligence built on a live context graph.
Backed by leading venture firms and deployed with Fortune 500 companies and leading enterprises, IndyKite supports how data is used across applications, AI systems, and digital ecosystems.
As AI agents and users execute work, IndyKite continuously captures context, relationships, trust signals, and actions. This becomes the point of decision, pulling signals from multiple systems and applying them at runtime, with full traceability.
The result is agentic AI that can operate across platforms with precision and deliver an entirely new class of automated intelligent services.
Learn more at IndyKite.ai