A production system where researcher, writer, and editor agents collaborate in Discord channels — scheduled automatically, routed to the right specialist, evidence-checked by the editor, and brand-aligned against your voice profile — turning one content strategist into a 10x output engine.
One strategist researches, writes, edits, formats, and publishes. At 3 pieces/week, quality holds. At 10, it breaks. Hiring writers adds coordination overhead. The factory model splits the work across specialized agents with built-in quality gates.
Research takes 2 hours. Writing takes 3. Editing takes 1. Formatting takes 30 min. That's 6.5 hours per piece. At 5 pieces/week, the strategist has no time for strategy — topic selection, distribution planning, performance analysis. The work that drives growth gets squeezed out.
The same person who wrote the draft edits it. They miss their own blind spots — weak hooks, unsupported claims, tone drift. A separate Editor agent with a different prompt (evidence checking, brand voice enforcement) catches what the Writer misses.
Add a freelance writer → you need briefs, review cycles, feedback loops, invoicing. Add two → you need an editor. The coordination tax eats the margin. Agents have zero coordination tax — they hand off via structured Discord messages with JSON schemas.
Define agent roles, set up channels, configure the scheduler, add AI quality gates, and connect the publish pipeline. One prompt deploys the full system.
Three outcomes: research depth increases because the Researcher only researches, brand voice stays consistent because the Editor only edits, and the strategist shifts from production to strategy.
Research returns the claim, the source and the date. Unsourceable claims are listed as a block at the end rather than quietly fixed, so a draft cannot hide its own gaps.
Three or four of your best posts are the reference. If a draft could have been written without reading the guide it comes back — and the agent says which rule each paragraph actually obeyed.
Weekly it names its own repeated failure — over-claiming, drifting, padding — so the guide gets fixed instead of the writer being told to try harder.
Are you ready?
All without you getting tasks done on manual, missing out the opportunity you want, or putting your hard work in vain.
Try it with Zaira nowThe practical questions people ask before building their agent team.
Blog posts, newsletters, LinkedIn articles, Twitter threads, case studies, white papers. Each is a Writer skill. Add new types by adding a skill to the Writer roster — no architecture changes.
Yes. For each claim in the draft, the evidence check searches the research brief (in Notion) and the web. Returns: SUPPORTED (source found), UNSUPPORTED (no source), CONTRADICTED (source says opposite). The Editor sees a claim-by-claim report.
The brand voice profile (Notion page) becomes the reference corpus. Each paragraph of the draft is compared against the profile's style examples and forbidden phrases. Returns a tone score 0-100 and specific violations (e.g., 'paragraph 3 uses forbidden phrase: unlock').
Yes. Create one factory per client: separate Discord server, separate brand voice profile, separate content calendar. The agent prompts are identical; only the configuration (profile, calendar, CMS credentials) changes.
The Editor's evidence pass catches unsupported claims. If a claim is central to the piece, the Editor marks REVISE with 'missing source for: [claim]'. The Writer re-hands-off to Researcher for that specific gap.
Track: pieces/week, avg time-to-publish, quality score, organic traffic per piece, leads per piece. The factory pays for itself when (pieces × traffic × conversion) > (strategist cost + Zaira plan).
Discord gateway connection