Accountable social publishing for AI agents

Give the agent a publishing lane.
Not the keys to the building.

PostSteward turns an agent’s proposed social post into an exact, reviewable effect: explicit owner approval, a cancellation boundary, durable state and independent provider readback.

The walkthrough below is deliberately non-effectful. Real account authority and publication stay inside canonical PostSteward.

automated tests
hosted checks
Threads OAuth
live publish + readback

Agents are getting more capable.
External effects need to get more accountable.

A social API call is not just another tool invocation. It creates a public, durable effect under a real account identity.

PostSteward separates what an agent wants to do from what the owner actually authorises, then preserves enough evidence to inspect what happened afterwards.

Five controls between intent and “it’s live”.

Simulation only — no post is sent

poststeward / effect trace
    showcase://non-effectful

    Every important claim gets a receipt.

    Canonical revision loading…

    Loading versioned evidence…

    The showcase is the stage.
    PostSteward remains the product.

    Public layerpoststeward-showcasestory · evidence · demo · launch assets
    Canonical productAyobamiH/poststewardauthority · state · OAuth · publication · readback

    The engineering is real.
    The final live proof is next.

    The remaining headline acceptance gate is one controlled Threads publication through the canonical owner journey followed by independent readback. This showcase will not mark it complete early.