AI Marketing Agents: A Lead Agent With Specialist Sub-Agents
Postellite's AI marketing agents work as a team on each of your websites: a lead agent that can use every tool and skill, plus specialist agents built for one job each. Long jobs are split across AI sub-agents that work in parallel, and you can watch every step as it happens.
A lead agent plus specialist agents by role
Ask the lead agent anything, or go straight to the specialist whose job matches the request. Every agent works inside the same workspace and business context.
Lead Agent
The all-round agent. It can use every tool and every skill, so it handles mixed requests from start to finish.
Research & Publishing Agent
Researches a topic, writes a search-optimised article and publishes it to your site.
Content Refresh Agent
Finds posts slipping in search and updates them in place, on the same URL.
Search Console Agent
Reads Search Console and your content to find opportunities, then writes performance reports.
Social Agent
Turns articles into posts for each social network and schedules them on the calendar.
Visuals Agent
Creates featured images and social visuals, reusing your own photos first.
Meta Ads Agent
Reports on Meta Ads, spots wasted spend and drafts better creatives.
Content Planning Agent
Plans what to publish and when, based on search data and your goals.
Long jobs split across sub-agents that run in parallel
A request like writing a full article is not one long prompt. The work is handed to sub-agents, and each one shows its own live steps and timing in the chat.
Researcher
Gathers the sources, search data and business knowledge the article needs.
Planner
Turns the research into a structure, with headings and the points each section must cover.
Writer
Drafts the article from the plan, using your brand voice and workspace knowledge.
Editor
Checks the draft for invented links, uncited statistics, duplicate titles and repeated phrases.
Illustrator
Prepares the featured image and visuals, starting from your own image library.
Lead Agent
Brings the parts together, publishes after your approval and re-checks the live page against what you asked for.
From one request to a verified result
Build agents that match how your team works
The built-in agents are a starting point. Create your own, or adjust the built-ins for each workspace.
Your Own Agents
Give an agent its own name, avatar and instructions, so it behaves the way your team expects.
Allowed Skills
Choose exactly which skills each custom agent may use.
Start From a Template
Begin with an agent template instead of writing instructions from a blank page.
Customise the Built-Ins
Tune a built-in agent per workspace, and use Reset to default whenever you want the original back.
Skills
10 built-in skills, plus skills you write yourself
Skills are reusable playbooks the agents follow. Postellite ships with ten, and you can turn any of them off or customise it per workspace.
You can also write your own: describe when to use it, the trigger words and the steps or rules to follow. Approvals still apply to every skill.
A live step timeline instead of a black box
Every tool call appears as a step in the chat, so you always know what the agents are doing and why.
Live Step Timeline
Each tool call shows as a step. Sub-agents appear nested, with their own steps and timing.
Plan Widget
For requests with three or more steps, a live to-do list tracks what is done and what is next.
Structured Questions
The agent asks with clickable options, image choices, dates or numbers, plus an Other box. Skip a question and it uses its judgement.
Live Cloud Browser
Watch the agent open, read, click and scroll real pages in a panel. It hands control to you for logins, 2FA or CAPTCHAs and never types passwords.
Rich Result Cards
Articles with quality checks and a Google preview, publish results, Search Console and Meta campaign tables.
Chats That Keep Running
Leave the page and the work continues. Come back, reattach, or compact a long chat so it can carry on.
Approval before live changes
Agents that ask before they touch your live site
Before any write to a live site, the run pauses with Approve and Cancel. The same applies to a browser click that would publish, send, pay or delete something.
After a write, the agent re-fetches the page and compares it with what was asked, so a change that quietly saved as a draft does not slip through.
Shared workspace memory
One memory across web chat, Slack and ClickUp
The agents reach the same business memory wherever you talk to them. A request started in Slack or ClickUp knows what the web chat knows about that workspace.
Each turn brings together everything the agent needs to answer well and act safely.
Put a Whole Marketing Team of Agents to Work
Start with the lead agent, call in specialists when you need them and let sub-agents handle the long jobs in parallel. Custom agents are included from the Starter plan.
A desktop app with a local agent workbench is coming soon.