Ora Partners with MIT's NandaIndex to Build the Open Agentic Web

Imagine a bakery whose ordering agent runs on AWS while its agent card is hosted elsewhere. How does another agent find it - and keep finding it when the bakery changes providers?

The challenge extends to individuals whose agents are tied to email or account identities rather than domains. Most won't operate their own discovery infrastructure, yet they still need to be reachable.

NandaIndex offers a simple answer: a stable identifier that resolves to the appropriate discovery mechanism, even as the underlying infrastructure changes.

Think of it as a switchboard, not another central directory. Given an identifier, an agent can locate the right catalog, gateway, directory, or agent card.

For people without domains, the architecture also offers a path through verified account identities associated with trusted providers - while leaving important identity-binding questions open.

NandaIndex doesn't replace existing discovery infrastructure. It makes that infrastructure reachable across boundaries.

There's a fundamental difference between knowing who you're looking for and knowing what you're looking for:

• NandaIndex answers: “I have an identifier. Where do I go?”
• Agentic Resource Discovery (ARD) answers: “I have a task. What can help me accomplish it?”

If an agent already knows the bakery's identifier, NandaIndex can resolve it to the appropriate agent card, catalog, or directory.

But if the agent starts with a request like “find a bakery nearby that can prepare a sourdough loaf for pickup tomorrow,” it doesn't have an identifier yet. ARD gives it a standardized way to search by intent and discover capabilities that match the task.

The two fit together naturally because they share a foundation in AI Catalog, the open format for describing agentic resources. NandaIndex reuses AI Catalog entries for resolution, while ARD builds on the same representation to make resources searchable by task.

This isn't just theoretical compatibility. Ora Directory is a live reference implementation of ARD, exposing a standardized discovery interface that agents can use to search for capabilities.

The systems can work together without replacing each other. NandaIndex can point an agent to an ARD directory, while ARD can discover resources independently, without a known identifier or an additional resolution step.

The goal isn't one giant directory. It's an internet where agents can find their way across many directories.

At Ora, we believe the web needs to be rebuilt around agents as an increasingly important class of user. That conviction has driven our work helping shape the open standards of the agentic web, including co-creating MCP Apps and contributing to the Agentic Resource Discovery (ARD) specification.

Ora Directory is a reference implementation of ARD, bringing standardized, intent-based discovery together with real-world agent-readiness measurements.

Agents can discover businesses, products, services, and capabilities through a common interface, including MCP servers, Skills, and APIs. With npx ax generate-files, businesses can generate, review, and publish machine-readable discovery files without building a custom directory.

But discoverability alone doesn't guarantee success. An agent can discover exactly the right business and still fail when it gets there: a site that blocks automated access, documentation it can't interpret, an API it can't authenticate against, or a checkout it can't complete.

That's why Ora Directory pairs capability search with readiness evidence - from static checks and real agent runs to cryptographically verifiable attestations - helping agents choose resources they can actually use.

In AX is the New AEO, we studied 37,927 agent journeys across 1,056 real businesses using four independent agent harnesses.

Agent-ready businesses were 1.9× more likely to be clearly recommended. The study measured agent readiness rather than NandaIndex, but shows why discovery must lead to usability.

Discovery tells an agent where to go. Agent readiness helps determine what it can accomplish once it gets there.

We're approaching the agentic web from complementary directions.

NandaIndex addresses a missing layer between identity and discovery, connecting fragmented systems through stable identifiers and federated resolution. Ora focuses on open capability discovery and the practical question of whether businesses and their capabilities are usable by agents.

Both are needed for an agentic web that works beyond individual platforms and enterprise boundaries.

As a NandaIndex launch partner, we're excited to work with the NANDA team and the wider community to advance three areas:

• Interoperable resolution and discovery. Making ARD discovery endpoints resolvable through NandaIndex using shared AI Catalog formats, so agents can move naturally from identifier resolution to capability search, without proprietary integrations or dependence on any single platform.
• Making the long tail discoverable. Agent discovery shouldn't be reserved for enterprises with infrastructure teams. Ora Directory and our publishing tools make participation simpler, while NandaIndex provides a path for small businesses and individuals to maintain discoverable identities across providers, even without their own discovery infrastructure.
• Discovery grounded in real agent experience. Bringing agent-readiness evidence into discovery, so agents can evaluate not just which resources match their intent, but which ones they can actually access, understand, and use. Ora's measurements and signed readiness attestations provide a foundation for exploring this across interoperable directories.

NandaIndex v0.1 is an early architecture proposal and request for comments. Important questions around identity binding, trust, and federation remain open. These are exactly the questions the ecosystem needs to solve together, through open standards, interoperable implementations, and real-world experimentation.

Our partnership reflects a shared belief: the agentic web should be built as an open ecosystem, not another collection of walled gardens.

The early web became transformative because anyone could publish a page, link to another, or build a browser. No single company owned the experience.

The agentic web deserves the same foundation. No business should need a proprietary platform to become discoverable, no individual should need to own a domain to have a personal agent, and no agent should need custom integrations with hundreds of registries.

We don't need to retrofit yesterday's search infrastructure and call it the agentic web. We have an opportunity to design discovery around how agents actually work.

NandaIndex provides a path toward connecting the discovery systems we already have. ARD gives agents a common language for searching across them. And Ora is working to make sure that what agents find, they can actually use.

The agentic web won't be defined by what agents can find. It will be defined by what they can get done.