features
Everything between “here’s my URL” and a published post
Magic Share runs the whole pipeline — research, writing, fact-checking, scheduling, publishing — and keeps you in the approval seat the entire way.

research & drafting
Drafts grow in a greenhouse, not in production
Add a URL and the agent scans your site into a brand profile — product, audience, tone, keywords. Each run it studies the niche, shortlists six to ten scored post ideas, picks the strongest, and writes a 600–4,000-word draft — length matched to the query — with internal links, plus a hero image when you ask for one. Everything stays a draft until you say otherwise.
- site scan → cached brand profile
- 6–10 scored ideas per run
- 600–4,000-word drafts sized to the query, hero image on request

verification
Every link checked before you read a word
Before handing a draft over, the agent checks each claim against its source and fixes or cuts any link that fails, and the app rejects a post that fails its schema checks — title and description lengths, slug, tags, target keyword. The agent's per-link verification report ships with the post, so reviewing a draft takes a coffee, not an afternoon of fact-checking.
- every cited link reported: verified, replaced or removed
- schema-validated frontmatter & metadata
- verification report attached to every draft

publishing
Approved posts flow wherever you publish
Approval is the gate; distribution is your choice. Open a pull request against your static-site repo — merging the PR is the final publish step, so nothing lands without a human. Or, on a paid plan, tweet the post's thread to X and schedule standalone daily tweets. Or download the markdown and HTML.
- GitHub PR — merge is the publish button
- X threads + scheduled daily tweets
- .md / .html downloads

scheduling
A content calendar that fills itself
Turn on auto-draft and the agent wakes at 06:00 UTC, writes whatever your topic queue has due, and goes back to sleep — every morning on Starter and Pro, drawing from your monthly credits, approval-first, never publishing on its own. A topic ledger remembers everything already written or planned so ideas never repeat, a banned-topics list keeps it away from subjects you'd rather skip, and the publish schedule spaces posts out on your cadence.
- auto-draft at 06:00 UTC on Starter and Pro
- topic ledger deduplicates ideas across runs
- banned topics & publish schedule per site
Pick your model (Pro)
On Pro, a curated picker of tool-capable models — from GPT-5.4 Nano to Claude Opus 5 — with per-run cost shown up front. GPT-5.4 Mini is the default everywhere. You pick the main model; subagents inside a run may use others.
Nothing to install
The whole content engine runs on our infrastructure — no API keys to manage, no local setup. Open the web app and the agent is already awake.
SEO-shaped output
Posts arrive with slug, tags, target keyword, meta title and description already validated against length rules.
Async by design
A post is researched and written in one run of about 5–9 minutes; deep research (a topic search and a deep dive before the write) takes about 15–25. Jobs run in the background, and we email you when a draft is ready.
Spend guardrails
A monthly credit bank meters every run — unused credits roll over, up to 3 months' worth, and retries of failed runs are free — with per-run cost ceilings underneath, so an automated agent never surprises you.
Your voice, not a template
The brand profile keeps tone and terminology consistent with the site the agent studied — every post sounds like you.
See it run on your own site
Add your URL, watch the agent build your brand profile, and review your first fact-checked draft — your first three posts are free.