← Platform Modules

AI-Native Capability Layer

Coming Soon

Flag enableCapabilityLayer

Coming Soon. This module is part of the launch-to-live operating layer and is not yet enabled in the current release. The description below explains the intended behavior so you can plan ahead.

The capability layer is the spine that makes Podium AI-native rather than an app with a chatbot bolted on. Every capability is a typed, governed tool exposed once and consumed identically by the human cockpit and by agents. The same contract performs the same action, whoever calls it.

The dual-consumer test

A capability passes the dual-consumer test when the cockpit UI and an agent invoke the exact same tool contract to perform the same action. Two separate code paths for one action would mean two systems drifting apart; one contract keeps the human and agent surfaces honest with each other.

Typed read / propose / mutate tools

Each tool is classified by its effect. Read tools return information. Propose tools draft a change for human approval. Mutate tools change state and are never exposed to agents autonomously. Bought engines and libraries are wrapped behind owned tool contracts and never exposed directly, so the platform owns its boundaries.

One service, two protocols

The layer is served from a single gateway that speaks both REST (for the cockpit) and MCP (for agents and external harnesses). That is how an agent framework can drive Podium through the same governed capabilities a human uses, with the read/propose/mutate boundary enforced centrally.

Status note

This is foundational to the roadmap but is held off deliberately until a front-end surface is wired to invoke tools and the tool endpoint is behind authentication. It is documented here so the architecture is clear ahead of enablement.

Related