All agents

Blog Publishing Agent

An always-on SEO writer that researches, drafts, and ships a publish-ready long-form blog post on your schedule — no manual writing required.

The Blog Publishing Agent is an autonomous SEO content writer that runs on a schedule (or on demand) and produces one complete, 1,000–1,400 word blog post per run. It researches current angles, writes in your configured tone with proper H2/H3 structure and a meta description, and saves the finished post as publish-ready markdown in the agent's asset store, where you can review and download it. The goal is a consistent publishing cadence that compounds organic traffic without your team touching a keyboard.

What it does

Each run, the agent takes your topic/keywords and tone, researches the subject using web search and by reading a key live source, then writes a single original, well-structured long-form article (intro, H2/H3 body, conclusion, and a meta description line) tuned for SEO. When the draft is final it calls its publishing tool (publish_post) exactly once with the title and full markdown body, which stores a publish-ready post in the agent's R2-backed asset store keyed to your deployment; you review and download it from the dashboard. It is capped at 6 turns per run, so a run completes quickly (typically 1–3 minutes) and produces exactly one deliverable plus a short summary in your dashboard feed.

The value is a reliable, low-effort content engine: feed it a keyword theme once, set a weekly cadence, and the blog stays fresh. Because runs are metered by token usage and logged, you get a clear record of what was written and when. It is built for founders and marketers who know content matters but can't keep up, agencies producing for multiple clients, and SaaS/e-commerce teams building organic acquisition. Note: the agent itself does not directly push to a CMS — its only publishing path in the runtime is the markdown asset store, from which you export or hand off to your CMS manually.

How it works

Your Business
1
Receive run prompt
On schedule or on demand, the agent is handed a prompt built from your config: the topic/target keywords and the requested tone. The prompt instructs it to write and publish this week's post and not to ask questions.
2
Research the topic
It calls web_search to find current angles and search intent, then web_fetch to read the readable plain text of one key source page (https only, truncated to ~4,000 chars) for live facts.
3
Outline and draft
It builds a structured H2/H3 outline, then writes a single original 1,000–1,400 word article — intro, body, and conclusion — in the configured tone (professional, conversational, or authoritative).
4
SEO pass
It adds a meta description line and clear H2/H3 headings so the post is search-ready before finalizing.
5
Publish the post
When the post is final it calls publish_post exactly once with the title and full markdown body. The post is stored as publish-ready markdown in the agent's asset store (R2, keyed to your deployment) for review and download.
6
Log the deliverable
The run ends with one complete publish-ready post plus a short summary of what was written; token usage is metered and logged to your dashboard feed.
Outcomes delivered

Setting it up — owner / admin

  1. 1
    Deploy from the marketplace
    Subscribe to the Blog Publishing Agent, then open its deployment page at /dashboard/agents/[id]. This is where you configure, run, schedule, and review outputs.
  2. 2
    Review optional integrations
    Research uses built-in web_search/web_fetch, so no connection is required. The marketplace listing also surfaces optional connections (WordPress/Webflow/Ghost for publishing, Google Search and Google Analytics for research/measurement), but note these are not wired into this agent's runtime tools — by default every post is stored in the built-in asset store for one-click review.
  3. 3
    Fill the config fields
    Set 'Topic / target keywords' (textarea — what to write about each run, e.g. 'AI automation tips for small businesses'), 'Tone' (select: Professional, Conversational, or Authoritative), and 'Run frequency' (select: Daily, Weekly, or Monthly).
  4. 4
    Set the schedule
    The Run frequency field drives automatic runs. Weekly is the typical cadence; choose Daily or Monthly to match your content plan.
  5. 5
    Do the first run
    After saving config, hit 'Run now' to generate your first post immediately. The run finishes in ~1–3 minutes and the publish-ready post appears in the deployment's outputs.
  6. 6
    Review and route the output
    Open the generated post, review the markdown and meta description, and download it to publish manually or hand off to your CMS. Treat the stored post as a draft and approve from the dashboard before it goes live.

Using it day to day — your team

  1. 1
    Check the dashboard feed
    Each completed run drops a new entry with a short summary in the agent's feed on /dashboard/agents/[id]. This is where you see that this week's post is ready.
  2. 2
    Open the post
    Click into the run to read the full publish-ready markdown — title, intro, H2/H3 body, conclusion, and the meta description line.
  3. 3
    Review and edit
    Skim for brand voice and accuracy. The post is written in your configured tone; tweak wording or headings directly before it goes out.
  4. 4
    Download or hand off
    Download the markdown from the asset store to publish manually, or hand it to your CMS. The agent stores the post for one-click review and export rather than pushing to a CMS automatically.
  5. 5
    Adjust for next run
    If you want a different angle or cadence, update the Topic/keywords, Tone, or Run frequency in config — the next scheduled run picks up the change.

Use cases

Programmatic SEO
Feed a list of target keywords and run repeatedly to produce a clustered series of articles on related topics over successive runs (one post per run).
Thought leadership
Turn a rough angle or product update into a polished, on-brand long-form post in your chosen tone.
Always-on blog
Set a weekly schedule and the agent keeps the blog fresh with zero manual effort, producing one publish-ready post per run.

What to expect

  • One complete, publish-ready 1,000–1,400 word SEO post per run (e.g. ~4/month on a weekly cadence)
  • Each post includes an intro, H2/H3 structure, conclusion, and a meta description line
  • A consistent publishing cadence that compounds organic traffic over time
  • Hours of writing and editing time returned to your team
  • A dashboard record of every run with a short summary and metered token usage

Metrics to watch

  • Publishing consistency — posts shipped per period vs. your configured cadence
  • Run success rate and time-to-complete (target ~1–3 minutes per run)
  • Organic traffic and impressions on published posts (measured in your own analytics, since this agent has no built-in analytics integration)
  • Editing effort per post — how much human revision the drafts need before publishing
  • Token usage per run, to keep metered cost in line with output

FAQ

Will it sound like me?
It writes in the tone you configure (Professional, Conversational, or Authoritative). For tighter brand voice you can refine the topic/keywords per deployment.
Does it publish automatically to my CMS?
No — the agent's only publishing action (publish_post) stores a publish-ready markdown post in its asset store. There is no built-in CMS push in the runtime; you download the markdown and publish or hand it off to your CMS manually.
Can I review before it goes live?
Yes. Every run produces a stored markdown post you review from the deployment page before publishing — treat it as a draft and approve from the dashboard.
How long is each post and how many do I get?
One original post per run, 1,000–1,400 words with H2/H3 headings and a meta description. On a weekly cadence that's roughly four posts a month; switch to Daily or Monthly in the Run frequency field.
Does it actually research, or just generate text?
It researches: each run uses web_search to find current angles and web_fetch to read a key live https source (truncated readable text) before drafting, so posts reflect real context rather than only model knowledge.