Discovery first
Agents can search by intent, apply exact filters, and inspect why a capability matched.
Public discovery pilot
Make digital capabilities easier to find and trust.
IntentRoute is a public, machine-readable capability discovery layer for AI agents. The current pilot brings together sourced web.search capabilities, ARD-compatible discovery, provenance, runtime freshness, and correction rules.
What we are building
A trusted capability graph for AI agents.
IntentRoute starts with free discovery: deterministic search, structured records, clear sources, runtime freshness, and a reading path designed for software and people.
Sourced records
Catalog entries preserve provenance, review dates, uncertainty, and a correction path.
Resilient choices
Trustworthy discovery is the base; recommendation, routing, execution, and settlement are separately gated layers.
Development stage
Professional discovery pilot. Measurable foundations.
Available
Searchable capability records
A curated set of sourced web.search capabilities is available through AI Catalog, OpenAPI, search/explore/list, and matching human and machine forms.
Current focus: retrieval quality, dependable facts, freshness, provenance, and correctionsLearning
Observe the pilot honestly
Retrieval quality, delivery, voluntary outcomes, corrections, freshness, and missing data remain separate evidence classes.
No request is treated as proof of usefulnessBuilt for discovery
A shared surface for agents, builders, and providers.
AI agents
Search the public capability catalog and inspect evidence, freshness, access facts, and explicit unknowns.
Read the resource map →Developers
Integrate once through the published compatibility profile and OpenAPI contract.
Explore developer notes →Providers
See how digital capabilities are described, sourced, reviewed, and corrected.
Explore provider direction →Current status
An ambitious product, released through evidence gates.
Available: a searchable, public web.search pilot with human and machine-readable paths.
In progress: evidence-based observation of retrieval, delivery, bounded outcomes, freshness, and corrections.
Working roadmap: recommendation, routing, execution, and settlement advance only when pilot evidence justifies them.
View the roadmap