Short answer
These systems solve different layers. MCP is useful when a model host needs a consistent way to discover and call local tools, resources, or prompts. UAI-1 is useful when an exchange needs a portable record that another team, runtime, reviewer, or release process can validate later.
When to use MCP
Use MCP when the main problem is connecting a model-facing host to tools, prompts, resources, and local context. It is most natural inside a runtime or application boundary where a host decides what capabilities the model can safely reach.
When to use UAI-1
Use UAI-1 when the exchange itself needs to survive as evidence: profile, delivery expectations, trust context, provenance, integrity, registry alignment, and validation outcome.
| Criterion | UAI-1 | MCP |
|---|---|---|
| Primary boundary | Portable public exchange record | Local tool/model exposure |
| Evidence posture | Designed for durable validation and review | Designed for runtime capability access |
| Best fit | Cross-team handoffs, release evidence, provenance | Local tools, resources, prompts, host integrations |
| Reader question | Can this message be checked after it crosses a boundary? | Can this model safely access the right capability inside this host? |
| Use together? | Yes. A local MCP workflow can emit or hand off UAI-1 evidence when state crosses a public or organizational boundary. | |
How they can work together
A workflow can use MCP locally and emit UAI-1 evidence when state crosses an organizational, public, or review boundary.
Common confusion
- Do not treat UAI-1 as a replacement for every local tool call.
- Do not treat MCP as the durable public record for a UAI-1 validation claim.
- Use canonical UAIX pages for UAI-1 details and the MCP specification for MCP details.