456
Workflow

Build a Multi-Agent Content Team

Researcher, writer and editor agents under a supervisor, producing publishable drafts.

2 min readupdated 2026-08-01

/ quick answer

Split the job into three specialised agents with explicit handoff contracts and an editor that can reject and reloop. Researcher, writer and editor agents under a supervisor, producing publishable drafts.

Researcher, writer and editor agents under a supervisor, producing publishable drafts. The problem it solves: A single prompt asking for a full article produces generic, unsourced content that still needs a full rewrite. Split the job into three specialised agents with explicit handoff contracts and an editor that can reject and reloop. It runs in 6 steps, starting with researcher agent: search + scrape, returns a structured brief (claims, sources, angles). This workflow node is part of the Onexial knowledge graph and links to related concepts, workflows and tools below.
Problem
A single prompt asking for a full article produces generic, unsourced content that still needs a full rewrite.
Solution
Split the job into three specialised agents with explicit handoff contracts and an editor that can reject and reloop.
Steps
  1. 01Researcher agent: search + scrape, returns a structured brief (claims, sources, angles).
  2. 02Writer agent: consumes only the brief, never raw pages. Outputs draft + heading map.
  3. 03Editor agent: scores draft against a rubric (accuracy, originality, structure, hook).
  4. 04Supervisor: if score < threshold, return the rubric notes to the Writer (max 2 loops).
  5. 05Human approves before publish. Log every rejection reason as future eval data.
  6. 06Track cost and time per accepted draft; kill any agent that never changes the outcome.
Tools Used
Prompts Used
Related Dictionary
/ frequently asked

What does the Build a Multi-Agent Content Team workflow do?

Split the job into three specialised agents with explicit handoff contracts and an editor that can reject and reloop.

What problem does Build a Multi-Agent Content Team solve?

A single prompt asking for a full article produces generic, unsourced content that still needs a full rewrite.

How many steps does Build a Multi-Agent Content Team take?

6 steps. It starts with researcher agent: search + scrape, returns a structured brief (claims, sources, angles). and ends with track cost and time per accepted draft; kill any agent that never changes the outcome..

Which tools does Build a Multi-Agent Content Team need?

It uses multi-agent-orchestration-stack, content-creator-stack — each linked below with its own node.