Changelog

What's new and improved in PostForge.

1.1.22026-10-29

New Features

• Article eligibility gate — topic discovery now proves each result is one published article before you can choose it. Publisher homepages, category, tag and archive pages, search results, feeds and digests are refused, and so is any page whose publication date falls outside your lookback window. Cards show when the article was published, and a refused page appears under Rejected with the reason in plain words and no Restore
• Weekly draft email — once a week, at the set hour in your own timezone, PostForge drafts a post from your best current article and emails it for approval. The email links back to the review screen with the reject reason ready to pick, and the Tips and weekly draft switch in Settings turns it off
• One Create Post screen — source, expertise, your take or brief, and the draft summary in a single stepped page
• Expertise from what you already have — drop a PDF or DOCX resume, or paste a profile page, and PostForge shows you what it understood. Topics are suggested only from what you confirm
• Restart the first-run wizard from Help

Improvements

• Drafts are no longer forced into one shape. The universal output example that demonstrated the same five-beat skeleton (opening, evidence, analysis, takeaway, close) in every draft is gone: free-form structure, the post template you choose, and your expertise brief each control the structure on their own
• Your post length is a budget, not a quota. A draft that makes its point in fewer words is accepted instead of being padded to hit a number
• Generic phrasing is flagged, not rejected. A post that merely used an AI-sounding word now goes to review with the phrases named, and a technical term you would actually use stays in the draft. Drafts are still held for what cannot be reviewed into correctness: the length ceiling, plain text, and the punctuation rules you set
• Authorship boundary: a draft never claims an experience you did not supply. First-person judgment requires your take, brief, or saved position, and no client, conversation, or past result is invented to make a post feel personal
• The agent tools carry the same source evidence — each candidate shows its publication date, a refused page comes with the reason, and asking to draft from a refused source now answers with an explanation instead of a draft
• The first-run wizard no longer asks for your voice profile — voice and editorial preferences live in Settings
• Discovery asks for article permalinks and event-oriented queries instead of generic "tools best practices" suffixes. Search excerpts and extracted full text are tracked separately, a source that was only ever excerpted is re-read when you pick it, and drafting refuses a source it cannot read instead of writing from a snippet
• The discovery cache now carries the eligibility policy version, so pages approved under an older policy are revalidated rather than reused
• The landing page leads with one audience label and the example post; the aggregate proof counts are gone

Fixed

• Onboarding step 2's search card shows a real loading state while it runs
• The expertise source form puts the file first and clears the other source when you drop one

1.1.12026-10-23

New Features

• Start from your expertise — pick the kind of post (a pattern, a decision rule, myth vs reality, a client question, or a framework), answer five questions about who it affects and where it stops applying, and PostForge finds a source that serves the brief and drafts around your position. Candidate cards say how well each source serves the brief, and expertise runs always pause for your approval
• Editorial preferences — set parentheses, dashes, hashtags, paragraph style, structure, and closing style once in the voice setup flow, and every draft follows them
• A voice step in onboarding — add writing samples or start from a preset before your first run
• Optional "Improve this post" tips at the approval step point at voice setup, topics, and expertise mode. They never change the draft you are reviewing, and closing them keeps them closed
• The MCP server gains a source-discovery tool for briefs, creating a run accepts an expertise brief, and the publishing skill covers the expertise-first path
• A public MCP guide at /mcp-docs/ — connect steps, the endpoint and a copyable Claude Code config, and every tool grouped by workflow, generated from the server's own registry. Linked from the footer and from Settings → MCP access
• New resource: Is LinkedIn Worth Your Time? Decide Before You Build a Posting Habit

Improvements

• Claim checks now name what each claim was compared against — the selected source or your expertise brief
• The landing page states the division of labour ("Bring the point. We'll handle the rest."), names who does each workflow step, and uses a new example post. Features and FAQ copy now match what ships: review is the default, scheduling runs inside the window you choose, and expertise, take, scorecard, claim check, and MCP are described where they belong

1.1.02026-10-19

New Features

• Claim check at the approval step — on paid plans, compare the draft's factual claims with the article you selected. Each claim is marked supported, contradicted, not established, or insufficient, with the deciding source passage. The check runs on demand against your current text and never blocks approval; trial accounts see the control disabled with a paid-plan note
• The PostForge MCP server gains a claim-check tool, and the publishing skill offers the check once with the review options before approval

Improvements

• The draft scorecard still does not verify claims. The claim check is a separate comparison with the selected source, and both the review panel and the agent skill label it that way
• SubQ (Subquadratic) is now a subprocessor for claim checks; the subprocessors page and privacy policy list it, its no-training statement, and its retention window

1.0.152026-10-14

New Features

• Your take — tell PostForge what the post should argue, in your own words or from suggestions drawn from the article, and it leads the draft. Save takes to a library, and set a default take in Voice & Brand for automated runs
• Source-first create flow — browse and preview candidate articles before starting a run, then start from the article you pick. A pasted URL skips discovery and begins at the draft step
• Plans now state a post allowance — 15 posts a month (45 quarterly, 180 annual). Unused posts roll over (up to 15), top-ups cover a busy cycle (5 posts, $6.99), and a one-time flexible pack (10 posts, $19.99, valid 12 months) suits occasional publishing
• Coupon codes can be applied at subscription checkout

Improvements

• "Unlimited posts" is gone everywhere it was claimed — pricing, FAQ, and comparison pages now state the allowance, and the billing page shows posts left this cycle, the refill date, and anything rolled over. Existing subscribers keep their current terms
• Enabling automated scheduling now requires an active paid plan; disabling stays available to everyone, and existing schedules keep running
• The draft scorecard now says plainly that its checks cannot verify claims. Review names, figures, and recommendations before publishing
• The landing page leads with "Source-grounded LinkedIn thought leadership" and got a calmer type and spacing pass; its draft angle is labelled a suggested interpretation unless you add your own take
• Comparison pages are generated from one data source, and each Compare hub card summarizes the difference
• New resources: why a source summary is not a LinkedIn post, and how the Elevate program helped PostForge move from a working product to a clearer business — plus an FAQ answer to "I can just use Claude or ChatGPT"

Fixed

• Drafts and take suggestions read the complete article — the extraction cap that cut sources off mid-article is removed
• A run started with no topics is refused up front instead of failing partway through
• URL runs skip the article-scoring stage and start their progress view at the draft step instead of replaying discovery
• Removing your last enabled discovery source is now explained inline instead of silently leaving an empty search
• App links in emails (like /create) now route signed-out visitors through sign-in instead of a dead view

1.0.142026-10-05

New Features

• Start your first post from an article link. Step 1 of the setup wizard accepts a public article URL as the source, so your first draft can come from something you already want to write about

Improvements

• The draft review shows your post once instead of twice. The separate copy with highlighted phrases is gone, and each flagged item is now a button that jumps to and selects that phrase in the editor
• Flagged items dim to "Removed" as you fix them, so the list matches your draft instead of repeating findings you have already dealt with
• The setup wizard says what each step will do before you commit to it: the step 1 note adapts to what you have entered, and the button reads "Continue with this link" once you have pasted one
• A run started from a link no longer claims an article was selected — its progress step reads "Using the link you provided"
• The second wizard step now says that most first runs finish within a few minutes

Fixed

• A draft held for your review no longer opens as an empty box. The post is saved before the review appears, so it always shows up instead of a blank panel that a page reload fixes
• A draft the model returns empty now fails with a retryable error instead of waiting for your approval as a blank post

1.0.132026-09-30

New Features

• Draft scorecard gains three judgements from the originality check: defined reader, added value over the source article, and whether a respected peer could reasonably disagree
• Drafts now address a defined reader — the voice profile's audience description is passed to the drafting model, so posts speak to a specific person's decision instead of everyone
• Draft validation flags more AI-writing tells: hollow intensifiers, overused consulting vocabulary, summary transitions, and repeated "not X, it's Y" frames. Each finding is explained in the review panel
• A banned word that is literally one of your topics (for example "thought leadership") may now appear in drafts as subject matter instead of being paraphrased away

Improvements

• Drafts aim at expert judgment instead of engagement mechanics: the prompt asks for a takeaway worth attaching your name to, a qualification or boundary drawn from the source, and an opening that can be a careful distinction rather than a contrarian claim
• Each post template now names the judgment it must carry, not just the topic angle

Fixed

• The draft scorecard no longer crowds the approve buttons on the review step

1.0.122026-09-23

Based on user feedback

• Publishing windows. Choose the days and the time range your posts go out in, instead of one fixed time per day
• Post templates. Choose the angle for your next post before you run it

New Features

• Early access — new accounts get 10 free LinkedIn posts instead of 3. The allowance is a one-time lifetime grant, existing accounts keep their limit, and the landing page states the offer and its conditions up front
• Publishing windows — pick one to six weekdays and a From/To range (overnight ranges such as 22:00–02:00 work), and PostForge prepares one post per selected day with a single random publishing time inside that window
• Post templates — Source to Insight, Adversarial, Research Lens, Product Promotion, Pattern Recognition, or Changed Mind. Available in Create Post and in the mobile app
• Approve for scheduled publishing — a prepared post waiting in review can be queued for its window or published immediately, as two explicit choices
• Posts now has a Scheduled tab that separates posts queued for their window from drafts kept after a missed window
• Draft scorecard — once a post is drafted, the approve step shows how it was judged: topic and audience relevance, insight, actionability, novelty, evidence quality, promotional risk, and an overall verdict

Improvements

• The landing page proof line shows live numbers — posts published, industry articles analyzed, and professionals using PostForge — instead of a fixed example count
• The landing page hero, workflow and proof sections quote the same real published post, badged as published unedited, and feature cards lead with outcomes instead of feature names
• The landing page adds early-user testimonials with customer logos, credited by name, role and company
• Pricing copy follows the 10-post early-access trial everywhere it is stated
• A window that closes before a post can be sent no longer publishes late: the draft is kept and labelled "Window missed — kept as draft", ready to publish manually
• Pausing and resuming automation preserves your days, window, and platforms instead of resetting them
• The dashboard and run banner show "waiting for its publishing window" for a prepared post instead of a spinner
• Fixed-time schedules convert to a window the first time you edit and save them; nothing changes until you do

Fixed

• The guided tour brings each step into view. Steps appear inline below the setting they describe, which left three of the five off-screen, so "Next" looked like it did nothing
• Contact now lives at /contact, the address people actually type — the old /contact.html redirects there. Before, /contact fell through to the app shell, so search engines treated it as a duplicate of the home page and left it out of the index
• Windows are interpreted in your display timezone, including the days the clocks change, so a window never silently shifts or disappears
• Changing your timezone or publishing days cancels automatic publishing for already-prepared posts; their drafts are kept and can still be published manually
• Deleting or cancelling a prepared post no longer lets the scheduler prepare a duplicate for the same day

Removed

• The Calendar page and its navigation entry. Everything it showed lives in three other places: your schedule (days and time window) is in Configure → Automation, Overview shows whether automation is active, and the Posts → Scheduled tab lists posts queued for their window plus drafts kept after a missed window
• The Content preference setting under Configure → Topics (Balanced, Research-focused, Product-focused). The angle of a post is now chosen with the post template in Create Post instead

Security

• Token encryption is enforced outside local development — the API refuses to start rather than store LinkedIn tokens and writing samples in clear
• Cross-origin access is restricted to an explicit allowlist, admin log errors no longer echo internal detail, and OAuth state checks are constant-time
• Development mock services now bind to localhost only

1.0.112026-09-18

Based on user feedback

• Custom discovery sources. Add urls for blogs, rss feeds as custom sources for articles

New Features

• Choose your discovery sources — switch the built-in web search, news feeds, and custom domains on or off, and see exactly what will be searched in Create Post before a run starts
• Install PostForge as an app — add it to your device straight from the browser
• Dismiss a failed run from Needs attention without leaving the dashboard

Improvements

• Landing page redesigned around one action: create your first free LinkedIn post. Six focused sections, one clear button, and a workflow you can explore step by step
• Broader audience — PostForge now speaks to busy experts across industries, with examples spanning technology, healthcare, and more
• Pricing leads with 3 free posts; all billing options sit behind one simple disclosure
• Coaching, competitor tables, the platform roadmap, and the long feature list moved off the main page to the footer and dedicated pages
• Mobile: Topics is now called Discovery

1.0.102026-09-08

New Features

• Upload your own post visual — choose "Upload my own image" in Create Post instead of an AI-generated one. JPEG/PNG/WebP up to 10MB is validated and auto-cropped to the same square format as AI visuals, so it flows through review and publishing identically. Available on all plans, including trial
• Purchase history — see every PayPal purchase and its status right on the Credits & billing page
• More honest article selection — candidates are scored in parallel with live progress, and unscored candidates show in their own "Not scored" group instead of disappearing

Improvements

• Checkout is now self-healing — returning from PayPal activates your subscription immediately, even if the webhook is late or misconfigured
• Annual plans are no longer charged monthly — the interval bug behind $99.99 monthly charges is fixed and self-corrects on your next checkout

Fixed

• Some purchases captured payment without activating the subscription or showing a confirmation — they now activate automatically
• Uploaded-image preview showing a broken image icon

1.0.92026-08-30

New Features

• Read-only maintenance mode — server-enforced guard during deployments with a non-dismissible maintenance banner in the app
• Blue/green production deployment — zero-downtime releases with validated graceful Caddy reload, automated drain, and instant rollback

1.0.82026-08-30

New Features

• New public pages — Features, How It Works, Pricing, and FAQ, linked from a shared header on the landing page
• Approval request emails — you're notified when a scheduled run needs your review
• Cleaner branded email templates — all transactional emails now render from a centralized template system with the PostForge banner
• Voice activation feedback — a persistent success message plus an active-profile summary with a quick Edit button
• Failed runs tab in Posts — see what went wrong instead of a silent gap
• Smarter article discovery — social media links are filtered out before scoring, and already-published or blocked articles are excluded
• Subscription expiry reminders — tiered emails before your plan renews

Improvements

• More reliable background publishing — schedule configuration is validated before every deploy, restart counts appear on the admin Diagnostics tab, and a systemd monitor alerts on crash loops
• Clearer error messages when AI draft generation fails

Fixed

• Background worker crash loop caused by an invalid schedule configuration — now caught before deploy
• Missing environment variables for the background worker (image generation, token encryption, scoring model) — now passed through
• Failed runs missing from the Posts list

1.0.72026-08-23

New Features

• User guidance system — every setting in Configure now has an inline help icon (ⓘ) explaining what it does and how it affects your posts
• Searchable help panel accessible from the ? button in the top navigation, covering all configuration options
• One-time guided tour for new users walking through key settings and their impact on post output
• Advanced image controls — Composition dropdown, plus Mood, Style, and Layout energy overrides in a collapsible Advanced section
• Custom image instruction field for fine-grained control over generated images
• Graduated onboarding reminders and low-credits warnings with smart cadence scheduling

Improvements

• Content Preference moved to Topics page where it belongs with other discovery settings
• Post Length moved to Voice & Brand for a more logical grouping with writing and visual controls
• Voice & Brand page restructured into three focused sections: Writing Voice, Post Length, and Visuals
• FAQ expanded with entries covering Content Preference, Writing Voice, AI images, publishing modes, and the help system

1.0.62026-08-20

New Features

• User communication emails — 24-hour "finish your first post" reminder for new users, and low-credits warning when your balance drops below 3

Improvements

• Workspace redesign — the app is now a focused content operations workspace with persistent sidebar navigation and dedicated pages for Overview, Create Post, Posts, Calendar, Topics, Voice & Brand, Automation, Credits, Referrals, and Settings
• Overview dashboard with automation health status, attention items, and recent activity at a glance
• Create Post workflow with transparent credit-cost display and progressive steps
• Posts page with filter tabs and per-post actions for managing your content
• Automation page with clear publishing mode controls and auto-publish warning
• Responsive mobile navigation with bottom nav bar and accessible drawer
• LinkedIn profile link on the Contact page for direct connection

1.0.52026-08-15

New Features

• Writing Voice profiles — create preset or custom voice profiles with sample-based AI analysis for personalized drafts
• Voice learning — paste writing samples and PostForge analyzes your style (tone, structure, vocabulary) to inform future drafts
• Validation issues tracking — runs now record specific validation issues for better visibility into post quality

Improvements

• More reliable article discovery pipeline
• Draft generation incorporates voice profile analysis when available
• Validation pipeline provides more detailed feedback on post quality

1.0.42026-08-14

Changed

• Landing page hero now leads with outcome messaging for consultants, founders, and technical professionals
• Pricing, FAQ, and comparison copy unified: a credit is only permanently used when a post publishes successfully

1.0.32026-08-04

New Features

• Image Theme dropdown — choose Corporate, Fun, Editorial, Futuristic, Minimal, or Bold art direction for generated post images
• Optional Accent Color — apply a brand color to generated images for visual consistency
• Native color picker — choose your accent color visually instead of typing hex codes
• Smarter prompts — scene type is auto-classified per article, and prompts use themed art-direction blocks
• Circular progress indicator shown over the image while an alternate is being regenerated
• Refreshing the page mid-run no longer resets the progress view — runs show the exact stage they're in

Improvements

• Faster image generation — images are delivered more quickly
• Even quicker re-runs — article scores are remembered along with discovered articles, so scoring is skipped when they're reused

Fixed

• Image regeneration no longer times out
• A run could fail with "No high-quality articles found" when reusing previously discovered articles — resolved
• Clicking an alternate article after a browser refresh now continues the workflow correctly

1.0.22026-08-02

New Features

• Image gallery — generate up to 2 alternate images in the draft preview and navigate between them with arrow buttons; whichever image is visible when you approve is published
• Published posts now load 10 at a time with a "Load more" button for faster page loads

Improvements

• Smaller page loads — frontend code is now split into separate chunks so returning visitors load less data

1.0.12026-07-28

Improvements

• New users now start with 10 free credits (up from 5)
• Sharper, higher-quality AI post images
• Stronger account and payment security

Fixed

• The landing page now fits phone screens cleanly — sample posts, steps, and footer no longer scroll sideways or misalign
• On/off toggles now clearly show whether they're enabled

1.0.02026-07-27

New Features

• Optional AI images for your posts — generate an accompanying illustration, preview it before publishing, and regenerate or remove it on the fly (image posts use +2 credits)
• Faster re-runs — recently discovered articles are remembered for 3 hours, so starting another run skips the wait for discovery
• Article rejection — trash an article you don't want to hide it from all future runs, and restore it anytime from the new Rejected tab
• Cancel a run at any point — stop a run in progress and start over, with credits refunded automatically

Improvements

• More reliable runs — a run interrupted by maintenance now recovers on its own instead of getting stuck, and never publishes twice
• The Discover step now shows the total time across all your topics rather than resetting on each one

0.9.42026-07-24

New Features

• Edit your draft before publishing — drafts now appear in an editable text box so you can refine the wording, and your edits are what get published
• New "Your presence at a glance" dashboard panel: posts published, weekly posting streak, words generated, topics covered, and last published date
• Landing page now showcases real posts written and published by PostForge, each linking to the live LinkedIn post

Improvements

• Comparison pages (vs Buffer, Publer, SayWhat) are now viewable while logged in
• Clearer, more accurate messaging about how PostForge keeps your posts on-brand

0.9.32026-07-20

New Features

• Runs now recover gracefully when the browser is closed and reopened — article selection and draft state restore automatically
• Clearer status indicators for paused and disconnected runs

Improvements

• Banner image optimized for faster page loading

0.9.22026-07-20

New Features

• New "Why automate" section on the landing page with value proposition cards
• Article URLs in the selection modal are now clickable links

Improvements

• Simplified fallback model chain for more reliable AI draft generation

0.9.12026-07-20

New Features

• Cerebras AI integration — gpt-oss-120b as primary model for ~3000 tokens/sec draft generation
• Automatic fallback to backup providers if primary model is unavailable
• SEO foundations: sitemap, machine-readable index, and enhanced social sharing metadata

Improvements

• Faster, more reliable AI drafts with the new primary model
• Run progress display and topic editor refinements
• FAQ content updates

Bug Fixes

• Video subtitles now embed reliably across all file path formats
• Pipeline recovery controls for resuming interrupted runs

0.9.02026-07-20

New Features

• Email verification — change your email and receive a verification link to confirm ownership
• Draft preview rendered as a non-copyable PNG image to protect your content
• Article full-text extraction for richer, more informed AI-generated posts
• LLM-based quality scoring for custom URL articles
• Confirmation emails sent after successful publish (opt-in per schedule)

Improvements

• Credits now deducted only after successful publish — failed runs no longer consume credits
• Approval buttons hidden during auto-publishing runs for a cleaner experience
• Confirmation email toggle moved to schedule edit view to reduce clutter
• LiteLLM timeout tuned so fallback models engage faster when the primary is slow

Bug Fixes

• Special characters (dashes, quotes) no longer show as broken symbols in draft preview
• Publish-complete email now has proper spacing between the article URL and hashtags

0.8.02026-07-18

New Features

• Color-coded article scores in discovery modal — green for highest, red for lowest
• Draft preview now shows word count in the section header
• FAQ entry explaining that posts can be edited on LinkedIn after publishing

Improvements & Fixes

• Discovery modal now shows top 12 highest-scoring articles
• Post length options refined: Short (75) to In-depth (375) words
• Faster, more reliable AI draft generation with updated models
• Pricing panels now display neatly in a single row
• Draft formatting improved — sections flow as paragraphs instead of one sentence per line
• Credit refund now works correctly when draft generation fails
• Admin credit grant works reliably on repeated grants

0.7.02026-07-18

New Features

• Adjustable post length: choose from Short (100) to In-depth (500) words
• Contact form for privacy and support inquiries

Improvements & Fixes

• Word count validation now adapts to your chosen post length

0.6.12026-07-18

New Features

• FAQ expanded with data privacy and account deletion policies

Improvements & Fixes

• Settings labels now display fully without truncation

0.6.02026-07-18

New Features

• Content preference: choose Balanced, Research-focused, or Product-oriented article scoring
• Privacy policy and changelog pages

Improvements & Fixes

• Configuration section layout updated for better readability
• FAQ expanded with data privacy and account deletion questions
• Changelog simplified to highlight what matters to you

0.5.02026-07-18

New Features

• Custom URL input: paste your own article link instead of using auto-discovery
• Referral program: share your link and earn credits when friends sign up
• Manual article selection: pick which article to use from discovered candidates
• Re-draft: regenerate a post from a different article at no extra credit cost
• Landing page with pricing plans, how-it-works guide, and FAQ
• Settings now auto-save on each change — no Edit/Save buttons needed
• Browser timezone automatically detected for new users

Improvements & Fixes

• Updated visual design with a cleaner, more professional look
• Credit system: deduction happens when you run, automatic refund only if publishing fails
• Post timestamps now display in the correct timezone
• Payment processing and credit delivery made more reliable
• Credit transactions verified against your account to prevent misuse
• Simplified infrastructure — tokens and configuration managed directly
• Various bug fixes and performance improvements

0.2.02026-07-17

New Features

• Admin dashboard for managing users, credits, and system configuration

Improvements & Fixes

• Improved AI reliability with automatic retries and backup providers
• Credit handling made more robust to prevent duplicate charges
• Smoother step-by-step progress display during runs
• Topic count now shown as "X of 5"
• Multiple stability and error-recovery fixes

0.1.02026-07-12

New Features

• LinkedIn sign-in and post publishing
• AI-powered draft generation with quality checks
• Review and approve drafts before publishing
• Scheduled publishing (daily, weekly, or weekdays)
• Credit system with PayPal payments
• Auto-approve option for hands-free publishing