RAI for Institutions

Why run your own registration agency?

Institutions with significant research computing infrastructure may want their own RAI prefix for several reasons:

• Institutional identity — agents registered under 18.mit or 18.inl carry the institution's brand
• Internal governance — control your own suffix allocation, approval processes, and metadata requirements
• Regulatory compliance — some organizations need data sovereignty or audit trails that require operating their own resolver
• Scale — high-volume registries benefit from dedicated infrastructure

What a registration agency operates

As an RA, you operate:

1. A resolver — an HTTP service that resolves RAIs under your prefix to descriptors
2. A registry — a database of agents registered under your prefix
3. An allocation process — rules for who can register identifiers and how suffixes are assigned

The resolver must implement the RAI resolution protocol:

• HTTP content negotiation (JSON + HTML)
• Version pinning via ?version=X.Y.Z
• Standard response codes (200, 302, 404, 410)
• /.well-known/rai metadata endpoint
• Custom response headers (X-RAI-Schema-Version, X-RAI-Agent-Version)

The application process

To apply for a prefix:

1. Submit an application to CIFR describing your institution, your intended use, and your technical infrastructure
2. Demonstrate resolver compliance — your resolver must pass the RAI conformance test suite
3. Agree to persistence commitments — identifiers minted under your prefix must resolve in perpetuity, even if your institution ceases RA operations
4. Receive your prefix — allocated by the Foundation and registered in the global prefix registry

The formal application process will be published in Prefix Policy v0.2. In the meantime, contact us to discuss your needs.

Using CIFR's infrastructure

Not every institution needs to build a resolver from scratch. CIFR offers hosted RA services for institutions that want their own prefix without operating infrastructure:

• Hosted resolver — CIFR operates the resolver under your prefix, with your branding on the HTML landing pages
• Delegated registration — your researchers register agents through CIFR's platform, but identifiers carry your prefix
• Full managed service — CIFR handles everything; you get a prefix, a dashboard, and institutional branding

Contact institutions@cifr.org.in for pricing and details.

Technical requirements

Resolver specification

Your resolver must implement the following endpoints:

Endpoint	Method	Response
/{prefix}/{suffix}	GET	Descriptor (JSON) or landing page (HTML) via content negotiation
/{prefix}/{suffix}?version=X.Y.Z	GET	Pinned version descriptor
/.well-known/rai	GET	Resolver metadata (schema version, supported prefixes)
/health	GET	Liveness check

Descriptor compliance

RAI descriptors returned by your resolver must conform to the RAI Descriptor Schema v0.1. Required fields:

• rai — the full RAI in storage form
• schema_version — must be "0.1" (or current version)
• identity — name, version, version_history
• agent — at minimum description, inputs, outputs
• trust — tier and image_digest
• resolution — self URL
• status — visibility and deprecated flag

Uptime commitment

RA resolvers should target 99.9% uptime. Resolution is citation infrastructure — a broken resolver means broken citations. The Foundation monitors RA resolver availability and may reassign resolution responsibility if an RA consistently fails to meet uptime targets.

Current registration agencies

Prefix	Organization	Resolver	Since
18.cifr	Council for Independent Frontier Research	rai.cifr.org.in	April 2026

Your institution could be next. Contact us to discuss prefix allocation.