Postdom
Designing and building an agent-first social publishing product for a client, from proposition and UX through delivery, pricing and launch.
Product and UX Lead, client engagement · 2026
for agents to plan, schedule, publish and measure
short-form video platforms supported at launch
Context
Postdom is a social publishing product built for AI agents. It gives an agent one place to plan, schedule, publish and measure short-form video across TikTok, Instagram Reels and YouTube Shorts, while the person using it keeps control of accounts and approvals.
I designed and built the product end to end for a client in 2026, using AI throughout the work. The engagement covered the proposition, UX, technical delivery, pricing and launch.
Problem
AI agents can write a post and propose a calendar, but publishing is an operational system. Each social platform has its own authorisation, media rules, scheduling behaviour, failure states and measurement gaps. A chat response is not enough when work has to run later and every destination needs a reliable outcome.
Most social tools are dashboards designed for a person to operate. Postdom needed a machine-readable contract for the agent and a clear control surface for the person supervising it.
My role
I worked across product, design and delivery: defining the market position, mapping the agent and human journeys, designing the account and approval model, shaping the API and MCP contract, building the public and authenticated product surfaces, and taking the service through pricing and launch.
Product decisions
Approve a bounded plan, not every click.
A person can approve a defined batch with post caps, timing rules and account boundaries. That gives the agent room to work without granting unlimited access.
Keep credentials away from the agent.
Social authorisation stays on each platform's OAuth screen. The agent receives the connection result, never the password.
Fail clearly rather than publish late.
A missed time stops with a reason. It does not silently publish at the wrong hour or leave the agent guessing about what happened.
Do not turn missing data into zero.
Every metric carries an availability state, so the next plan is based on evidence rather than a number the platform never supplied.
Execution
The work moved from the core proposition into the agent contract, workspace and account model, design system, publishing workflow, measurement model, pricing and public launch. Product and engineering decisions had to be made together because the promise of dependable publishing rests on durable execution underneath it.
The product supports direct agent use, publishing embedded inside another SaaS product, and separate client workspaces for agencies. Human approval, autonomy limits, pause controls and attributed actions are part of the main flow rather than a settings afterthought.
Outcome
Postdom launched as a self-serve product with free and paid plans, agent keys, connected social accounts and publishing across TikTok, Instagram Reels and YouTube Shorts.
| Launch product | |
|---|---|
| Agent integration | API and MCP tools |
| Publishing destinations | 3 platforms |
| Control model | Human-approved boundaries |
| Measurement | Explicit availability states |
This case study describes the public product. Client commercial details are not published.
What I learned
The interesting part of an agent product is not the chat. It is the contract between intent and execution: what the agent is allowed to do, how delayed work survives, how a person can intervene, and how the result comes back without being invented.