Support
Log in Sign up
WordPress · AI content

Write more.Publish only what earns it.

Draft Forge mines topics out of the site you already have, drafts them in a voice read from your own published posts, and rewrites the ones that have gone stale. Then it reads the result the way a person skims it and refuses to publish filler — and an author can hand the draft to the editor who is allowed to say yes. With Pro the whole thing runs on a schedule that stops itself. Drafting costs credits, not a subscription.

being forged — launching with the Avakode suite
5 checksword count, internal links, an image direction, distance from your own archive — and one that reads the prose and quotes the sentence it objects to
Author → editora draft can wait for the person you gave the review capability to, and the schedule will not publish past them
3new posts a week is the autopilot’s default ceiling, and every tenth hands-off publication pauses it until you confirm
The gate

The model will write anything. The gate decides what you publish.

3 of these 5 drafts never reach the site. The gate wants 600 words, 2 internal links, an image direction, and enough distance from what you already published — and it is the same gate whether a human pressed the button or the schedule did.

2 of 5 drafts made it past the gate
  • 600 words minimum
  • 2 internal links
  • 0.85 max similarity to your own posts

Every threshold is a filter you can change, and a refresh skips the similarity check on purpose — a rewrite is meant to resemble the post it rewrites.

Voice · gate · review

It writes in your voice, then argues with the draft before you have to.

Topics mined from the site you already have, a voice read out of your own posts, five checks between the model and publish, and an editor who still gets to say no.

Topics out of your own archive

The heuristic pass finds the gaps in what you cover and makes no network call at all. The AI pass reads for intent instead of keywords, costs two credits and says so on the button. Every topic lands in a queue with its source marked.

It learns your voice from your own posts

“Analyse my voice” reads up to five posts you actually published — yours before the site’s — and comes back with your tone, sentence length, person, and the words that would read as somebody else. The summary is offered, never applied behind your back, and once you keep it every later outline and draft is written to it.

Five checks, and the fifth one reads

Length, internal links, an image direction and distance from your own archive are measurements. The fifth is a lint: 132 stock phrases counted as a density, four kinds of repetition, an introduction that only restates the title, a rhythm too even to be human. It hands back the sentence with the objection beside it rather than a score.

The author generates, the editor approves

“Request review” moves a queued draft to whoever holds the review capability — Administrator and Editor by default. They approve and publish, approve as a draft, or send it back with a note; both ends are recorded on the row. The reviewer is e-mailed at most once an hour, and the autopilot refuses to publish past a draft still waiting on them.

Read the rewrite before it replaces the post

A refresh replaces a whole live article, so nothing is applied unread: the proposal beside the original, block by block and word by word, compared against the body the model was actually handed — and the screen says so if somebody edited the post in between. Reject writes nothing at all.

An autopilot that stops itself (Pro)

Put the pipeline on a schedule and then leave it: at most three new posts in any rolling seven days by default, and a checkpoint every ten hands-off publications that pauses new posts going live until you confirm. Drafting, refreshes and topic mining carry on through the pause.

Two meters

You pay for the words once.

Most AI content tools charge a subscription and then meter the AI inside it. Here the two are separate on purpose: credits buy model calls and work on the free plugin; the subscription buys the schedule. Neither pays for the other.

Credits

The model

  • Topic mining, outlines, drafts, refreshes
  • Works on the free plugin, no licence needed
  • One wallet across every Forge plugin you own

You spend them when you generate, and only then.

Subscription

The autopilot

  • Scheduled generation and refreshes
  • Publishing without a human in the loop
  • Bulk topic queue

Flat per site, whatever the volume.

And the gate applies to both: 600 words, 2 internal links, an image direction and enough distance from your own posts — whether you pressed the button or the schedule did.

A first post

Mine a topic, read the objection, then hand it over.

01

Mine your own site

The first run reads what you already published and proposes what is missing from it. The same posts, read again, become the voice profile every later draft is written to.

02

Generate, then read the objection

Do a few by hand. When the gate holds one, the report names the check and quotes the sentence — which is how you learn the shape of a topic that works here.

03

Hand it over on your terms

Pro puts it on a schedule with a weekly cap and a checkpoint you confirm. Or drive the same pipeline from the command line, where every command prints its price before it spends anything.

The autopilot

Run one tick. Read what it did afterwards.

A cron tick pulls pending topics and stale posts, runs them through the same pipeline you would by hand, and — if you allow it — publishes what passes. Then it writes down what happened, which is the part that lets you leave it alone.

scheduled run licence: active gate: 600 words · 2 links · image · similarity < 0.85
  1. How our pricing works 678 words · 2 links published
  2. Our 2024 guide, brought up to date 678 words · 2 links · refresh published
  3. A quick note 72 words · 2 links too thin
run summary

2published

1held, with the check that stopped it

A refresh skips the similarity check by definition — a rewrite is supposed to resemble the post it rewrites. Everything else is the gate you set, and automatic publishing is off until you turn it on.

2 published, 1 held — and the run said which check stopped it.
Hands off to

Fits the rest of the suite

Rank Forge
Lang Forge
WordPress editor
WP-CLI
wp-cron
Forge credits
Comparison

Draft Forge vs a chat window and copy-paste

What you are comparing Draft Forge a chat window
Where the draft lands A real WordPress post, with the target keyword handed to Rank Forge A chat transcript, then a paste and a reformat
Where the topic comes from Mined from what you have already published Whatever occurred to you this morning
What reads it before it goes live Five checks, and the fifth quotes the sentence it objects to Nothing. You are the gate, every single time
Who is allowed to say yes The editor you gave the capability to — and the schedule will not publish past them Whoever has the tab open
Sounding like you A voice profile read out of your own published posts, carried into every draft A style guide you paste again at the start of every session
Leaving it running With Pro: a schedule with a weekly cap and a checkpoint that pauses until you confirm Nothing happens while you are asleep
What it costs Credits per generation — nothing when you are not writing A monthly seat whether you write or not
Steering one paragraph One prompt chain, then you edit in WordPress You can argue with the model turn by turn until it is right their win
Writing something nobody has covered It looks for the gaps in a site that already exists A blank page and a conversation, which is where genuinely new writing starts their win

If you publish twice a month and enjoy the writing, you do not need this. It earns its place when the queue is longer than your evenings.

Pricing

Credits buy the words. The licence buys an autopilot that stops itself.

No upsells, no surprises.

Personal
$39 /year
1 site
Get it on avakode.com
Most popular
Professional
$89 /year
5 sites
Get it on avakode.com
Agency
$159 /year
unlimited
Get it on avakode.com
Before you buy

What runs where in Draft Forge

Which work happens inside your own install, what leaves it, what needs a paid plan, and what a lapsed licence leaves behind.

What the free tier does inside your own install

  • Five checks before publish, and the fifth one reads
  • The author generates, the editor approves
  • The fifth check, and the only one that reads rather than counts. A body can clear every threshold above and still open with "In today's fast-paced world", promise to delve into its own title, and repeat one sentence shape forty times. The lint answers with findings instead of a score: a bundled list of 132 stock phrases measured as a density per 1000 words, four kinds of repetition, an introduction that only restates the title, a rhythm too even to have been written by a person, dates and figures that cannot be true, and text laid out as a list where it should be prose. Every finding quotes the sentence and says what is wrong with it, in the Queue tab and on the refresh review screen. Three strictness settings — lenient, standard, strict — decide which findings stop a draft; the checks themselves always run at all three. Free, and it costs no credits: the credits already paid for the draft

What leaves your site when a feature is on, and where it goes

  • Usage telemetry is off until you switch it on: every send reads the site option `forge_telemetry_consent` first, and stops when it is not "1".
  • With telemetry on, batches of events POST to https://api.avakode.com/telemetry/batch.
  • The in-admin feedback form POSTs your rating and comment to https://api.avakode.com/feedback when you press Send.
  • The Forge Suite dashboard health card GETs https://api.avakode.com/health.

What needs Pro, credits or an outside service

  • Pro is a separate plugin. With no Pro add-on loaded the plan is free, whatever is stored in the database.
  • Licence activation and every AI request go to https://api.avakode.com.
  • AI actions spend credits from your Avakode balance; the price per action is decided on the server (1 credit for a translation, 3 for a whole post).
  • A request over 10,000 tokens costs a multiple of the base action price.

What happens when the licence lapses

  • Free features keep working. The wp.org base ships no licensing SDK: without the Pro add-on answering "pro" through the `forge_pro_plan` filter the plan reads free.
  • When the paid ladder answers "free" in so many words, the stored grace marker `forge_license_last_pro_at` is cleared and Pro gates close on the next admin request.
  • Pro keeps working for 6 hours of silence from the licence server — that window covers an outage, not a revoked licence.
  • A lapse removes no data. Tables and options go away only when you uninstall the plugin.

How long data is kept, and how you take it out

  • Generated drafts are ordinary WordPress posts with no automatic pruning.
  • The credit-usage log behind draft generation drops rows older than a year via the shared daily cleanup job.
  • There is no export of drafts or generation history.
  • The only CSV-related feature in the plugin imports decline data from elsewhere, not the other way round.
  • Uninstalling the plugin drops its table.

The same five questions for every tool →

Switching

No import, no re-platforming, and nothing published behind your back.

Nothing to migrate

No import and no re-platforming. The first run reads the site you already have and proposes what is missing from it; read the same posts again and it tells you how you write.

Your editor stays your editor

Drafts are ordinary posts — Gutenberg, your blocks, your workflow. And if you want the human editor too: the author asks, the editor answers, and the row remembers which of them did what.

Nothing publishes itself unless you say so

Every draft waits behind the gate, and hands-off publishing is off until you switch it on. Once it is on it still stops — at the weekly cap, and at the checkpoint that waits for your confirmation.

FAQ

What people ask about AI content that publishes itself

Not silently, and no longer only by the numbers. Four checks measure the draft — length, internal links, an image direction, distance from your own archive. The fifth reads it: stock phrases by density, sentences and openings repeated, an introduction that only restates the title, a rhythm too even to be human. It hands back the sentence and the objection instead of a score, at whichever of the three strictnesses you picked, and it costs nothing — the credits already paid for the draft.
That is what the review workflow is for. An author presses Request review and the draft moves to pending; whoever holds the review capability — Administrator and Editor by default, and deliberately not Contributor — approves and publishes, approves as a draft, or sends it back with a note. Both ends are recorded on the row with the name and the time, the reviewer is e-mailed at most once an hour however busy the day is, and a scheduled run will not publish past a draft that is still waiting.
Four things, and they fail in different directions. The gate holds anything that does not pass. The weekly cap publishes at most three new posts in any rolling seven days by default, whatever you set per run — the extra drafts are still written, they simply wait. A checkpoint after every ten hands-off publications pauses new posts going live until you confirm on the Schedule tab, while drafting and refreshes to posts already live carry on. And over all of it sits a monthly credit ceiling you set yourself, which the add-on can neither raise nor lower.
Only if you leave the voice field empty. “Analyse my voice” reads up to five posts you actually published — yours before the site’s — and describes how you write: tone, sentence length, person, the words that are yours and the ones that would read as somebody else. Two credits, offered as a summary you are free to decline, and folded into every outline and draft after you keep it.
Two separate things, on purpose. Model calls are metered by Forge credits from the wallet every one of our plugins shares, and they work on the free tier: about six credits for a draft, three for a refresh, two for an AI mining run, two for a voice analysis. The subscription buys the autopilot — the schedule, hands-off publishing, and the cap and checkpoint that hold it back. The gate, the lint, the review workflow, the calendar and the command line are free.
Six WP-CLI commands cover the pipeline: mine, generate, refresh, gate, queue and credits. A terminal is the easiest place to spend money by accident, so every command prints what the run will cost before it runs, the ones that spend refuse to without an explicit yes, and any refusal exits non-zero. Gate spends nothing and exits non-zero on a body the gate would hold, which is what makes it usable in a build.
Keep it current

A site that keeps itself current.

The free core mines, drafts, refreshes, lints and routes a draft to its editor. Pro adds the schedule — and the two brakes on it.

Forge AI Assistant Online

Hi! I'm the Draft Forge AI assistant. Ask me anything about the plugin — setup, features, troubleshooting, or development.

Just now
Powered by Forge AI · Browse docs