Release notes

Highlights

• The AI assistant no longer goes silent on you. Every message now gets a response — either the answer, or a clear "something went wrong" notice with a Retry button right in the conversation. Previously, if a reply failed partway through, your message could simply disappear with no error and no way to recover. That dead-end is gone.
• When the assistant is working through a complex request, you now see a brief "still working…" note instead of wondering whether it stalled — and a background safety net makes sure you’re never left staring at a turn that silently died.
• Uploading a document the AI can’t turn into a diagram now gives you a visible explanation in the chat instead of a blank result.
• Diagrams that were stuck showing every shape piled in one corner now repair themselves automatically the next time you open or preview them — you don’t have to do anything, and a clean layout is saved for next time.

Fixes

• Your saved diagrams are now protected from being overwritten by a broken layout. If an automatic layout step produces a malformed result — every shape stacked in one corner, or a swimlane diagram that lost its pools — the app refuses to replace your good diagram with it, instead of saving over your work.
• Retrying a failed assistant reply no longer risks leaving a duplicate behind — a retry cleanly replaces the failed turn.

Highlights

• Have a promo code? You can now enter it directly inside the upgrade modal — click "Have a promo code?", type the code, and the discount is applied automatically when you proceed to checkout. No need to visit a special URL first.

Highlights

• The free monthly limit now counts only diagram generations — chat with the AI assistant and behind-the-scenes refinement steps no longer eat into your 3-per-month allowance. The promised "3 generations per month" now means 3 actual diagrams in the editor, not 3 total messages.
• Auto-retries no longer count against you either: when a diagram needs a couple of internal validation passes before it lands, that's still one diagram against your cap — not two or three. Production data showed an average diagram was silently consuming ~2 of 3 cap slots before this fix.
• Failed generations don't count anymore. If our AI hits a snag and can't produce a usable diagram after all retries, you keep your full free allowance — only successful diagrams reduce your cap.
• Clarification-only attempts also don't count. If the AI asks you for more details and you walk away without finishing, that "parked" generation no longer eats into your monthly quota.
• Practical effect: you can chat with the AI as much as you need, and only the diagrams that actually land in the editor count against your 3-per-month free quota.

Fixes

• The "Monthly Limit Reached" paywall no longer fires for users who haven't generated a diagram yet — it now appears at the point you've actually used your 3 generations, not after 3 chat exchanges.
• Diagrams that succeeded on the second or third validation attempt no longer secretly count as multiple generations against your monthly cap.
• Generations that ultimately failed (LLM error, validation rejection across all retries) no longer count against your monthly free allowance.
• Generations parked at a clarification question (no diagram produced) no longer count either.

Highlights

• Pay-as-you-go credit packs are now sized for real export volume: $5 buys 30 credits with a 60-day expiry. One pack covers about 30 watermark-free PNGs, 30 SVGs, 15 PDFs, or 10 BPMN XML exports (any mix). Credits unlock paid exports, not generations — no more guessing how many "uses" you bought.
• New pricing-page calculator: tell us how many exports/month you need per format and we show whether credit packs or Pro is cheaper for your workload, with the savings spelled out in dollars.
• Pro is now $19/month (was $15). Team is $15/seat/month (was $12). Existing subscribers are grandfathered at their original price as long as their subscription continues without interruption — you got a separate email confirming this. New customers: visit bpmnai.com/pricing?promo=LAUNCH20 before you upgrade to claim 20% off the first 3 months — the discount is stashed when you land and applied automatically at checkout.
• PNG with watermark stays free on every plan; removing the watermark costs 1 credit. SVG: 1 credit. PDF: 2 credits. BPMN XML: 3 credits. The dropdown shows the cost upfront so nothing is a surprise.
• Two new milestone emails: one prompts you with the credit-pack option the day after your first diagram if you haven't exported a clean version yet; one celebrates your first BPMN XML / PDF / SVG export with a Pro discount nudge.

Fixes

• /pricing page now has a 6-question FAQ covering credit cost, expiry, refunds, Pro upgrade carry-over, pack stacking, and the free PNG path — questions you asked before signing up are now answered before you click.
• The pricing-page comparison table, home-page hero, and text-to-bpmn FAQ are now consistent about which BPMN tools we export to cleanly: Camunda, Bizagi, Lucidchart, and Signavio in the primary list.
• Marketing-page export claims are versioned alongside the actual code that ships them, so the numbers can't drift from what the app actually does.

Highlights

• New pay-as-you-go credit packs: buy a small batch of generations ($5 for 2, $12 for 6) without committing to a subscription. Credits are used automatically when you run out of free generations, and expire 12 months after purchase.
• Your credit balance and each pack’s expiry date now show in Settings → Billing, with a reminder email before credits expire.
• If a generation fails because of a system or AI-provider error, the credit you spent is automatically refunded.

Fixes

• The free monthly limit now counts every AI action (generation, chat, and refinement) against one shared allowance instead of leaving chat and refinement uncapped, and the monthly allowance is now 3. The cumulative free-diagram limit is unchanged.

Highlights

• Feedback popovers no longer land on top of a diagram the moment it finishes generating — the first-diagram feedback prompt now appears briefly after you export your first diagram instead, so you get a chance to actually look at the result before being asked about it.
• The "why are you hesitating to upgrade?" prompt no longer shows up to people who hit a paywall before they have generated a single diagram.

Highlights

• Chat replies now start arriving several seconds faster on every diagram conversation — most noticeable on document-attached chats and complex requirements.
• Onboarding has three clear starting points: try an instant sample diagram, pick a starter template, or describe your own process. The same three doors mirror onto the dashboard "Next step" card so the path to your first diagram looks the same whether you sign up fresh or come back later.
• BPMN 2.0 exports are noticeably more conformant — loops and multi-instance activities, compensation, data objects, text annotations, and nested lanes now survive the round trip into Camunda, Bizagi, and other strict modeling tools.
• Stronger generation pipeline: a structural validator catches dual-actor and cross-pool collisions during retries, and prompt hardening reduces the dual-modeling and missing-actor patterns that produced unusable diagrams.
• Manual edits in the editor are now preserved when a diagram is saved and reloaded — your canvas tweaks no longer get silently reverted on round trip.
• Profile tab now surfaces the Role and Industry you set during onboarding (read-only) with a link back to the Context tab for edits, so those answers are no longer "lost".

Fixes

• Auto-layout: ghost-lane handling extended to chained remaps and tricky multi-pool patterns, the layout engine no longer trips on cyclic graphs, and runaway pool/lane widths are capped so a single bad diagram can no longer balloon to hundreds of thousands of pixels wide.
• Polished chat UX: light-mode chat bubble is now clearly visible (was nearly invisible on white), and refining a generated diagram on the free tier shows a clear "Pro feature" hint above the composer instead of a confusing error toast.
• Dashboard counters update right away when you create your first diagram from the "Start with chat" door, instead of staying at zero for up to 5 minutes.
• PNG export retries with a safe size cap when SVG dimensions cannot be parsed, so more exports finish successfully on the first try.
• Layouts are checked at save time and broken ones are flagged before they reach the editor.
• Chat input on the new-diagram page is now properly aligned with the rest of the layout.

Highlights

• Iterate on a diagram from inside the editor — a side chat lets you ask for changes and see them applied without leaving the canvas.
• Editor redesign: refreshed top bar and canvas chrome, a palette and inspector with tabbed properties, and a morph dropdown to switch task, event, and gateway subtypes in place.
• New /conversations page lists all of your past chats so it is easier to pick up where you left off.
• New /text-to-bpmn landing page for people arriving from search.
• Pricing page now matches the live plans, and free-tier users see a clear paywall before iterating on a generated diagram.

Fixes

• Graceful redirect when your session expires instead of a broken page.
• PNG export no longer fails on very large diagrams (capped to a safe maximum size).
• Chat-driven generation progress now settles correctly and the in-editor chat pane stays usable on smaller screens.

Highlights

• System / light / dark theme toggle across the app, with a contrast pass to keep light mode legible.
• Dashboard rebuilt as a chat-first home so you can start a new diagram without hunting for the right button.
• Self-service account deletion with a GDPR data export, available from your account settings.
• Redesigned marketing landing page.
• Diagrams that pass partial validation are now delivered as "completed with warnings" instead of failing outright, so you still get a usable draft to edit.
• A read-only chat drawer in the editor lets you see the conversation that produced a diagram while you work on it.

Fixes

• More resilient sessions: stale tokens stop knocking you out, and protected routes are now guarded server-side.
• Generation retries are smarter when validation fails, so more drafts make it through on the first try.
• A crash in the editor tour has been fixed.

Highlights

• Sign in with Google in addition to email — works for new signups and links to existing accounts.
• Onboarding is now a single screen, and email confirmation is deferred so you can start working immediately after signing up.
• Pricing restructured with a new Enterprise tier; the pricing page and CTAs are aligned with the live plans.
• Free-tier limits and upgrade prompts are clearer in-product so you know when you are about to hit a cap.
• Founder About page, plus an RSS feed and updated sitemap.
• Cookie consent banner so you can manage tracking preferences.

Fixes

• Diagram exports include a small watermark on free-tier accounts (paid plans can opt out), and the watermark is now applied consistently to dashboard list exports.
• Billing simplified to prevent duplicate subscriptions; the free-tier upgrade CTA now routes to billing settings instead of a dead end.
• Team plans now enforce a minimum of 2 seats, matching the published pricing.
• Search-engine indexing fixes for the marketing site.

Highlights

• Landing page and lifecycle email copy rewritten for clarity.
• Post-trial upgrade flow tightened so it is clearer what changes when your trial ends.
• Diagram generation handles a wider range of LLM responses, reducing flaky retries.

Fixes

• BPMN exports now preserve embedded boundary events, so timers and errors attached to tasks survive a round trip.
• Duplicate-node warnings are surfaced during XML generation instead of silently breaking the export.
• Backend hardening for fewer transient failures during long sessions.
• In-product surveys are back after a brief regression.

Highlights

• Annual billing is now available alongside monthly, with the free-tier trial updated to match.
• Promo codes can be applied at checkout for marketing campaigns.
• Free tier now counts diagram generations rather than seats, with post-trial monthly quotas and lifecycle reminder emails.
• Comments and reviews are gated to paid team projects, with edit-sharing aligned to plan features.
• One-click unsubscribe in marketing emails plus a notification-preferences toggle so you can opt out of marketing without losing transactional mail.
• Per-page titles across the app and an auth-aware top nav.

Fixes

• Auto-layout: better swimlane spacing, cleaner cross-lane vertical handoffs, sequence-flow clearance, and your manual layout tweaks are preserved across reloads.
• Realtime updates and sign-in stay reliable across long sessions.

Highlights

• Team plans now have proper invite and seat-limit enforcement, so you can manage who is on each team without surprises at billing time.
• Free-tier enforcement is consistent across the product, including clearer messaging when a limit is hit.

Fixes

• Billing and trial handling hardened; renewal tracking is now reliable so reminders fire at the right time.

Highlights

• Upload documents in chat and get a response once processing finishes.
• Upload progress now stays in sync so you always know when files are ready.

Fixes

• Chats stay responsive even when files are still processing in the background.

Highlights

• Subprocesses are now first-class: open embedded flows in their own view, edit them, and return to the parent process without losing work.
• PDF exports can now include the parent process plus each subprocess as separate pages for easier sharing and review.
• BPMN XML exports now merge subprocess definitions into a single file and include a short export note about message-flow routing for readability.
• Validation issues are surfaced directly in the editor (including subprocess views) so you can fix structural problems without blocking preview.

Fixes

• Improved handling of embedded subprocess lanes and flows so exports and validators are less likely to break on complex diagrams.
• Generation failures now show more consistent, actionable messaging when unresolved validation issues remain.

Highlights

• Diagram editor now includes accuracy and layout ratings with optional notes so feedback is captured per diagram version.
• Auto-layout now offers quick sketch and BPMN swimlane modes (with keyboard shortcuts), cleaner message-flow routing, and saves enhanced layouts plus quality stats back to your diagram.
• Strict generation retries with participant-specific guidance when validation fails, while the editor can re-run validation or rebuild a draft from your saved diagram so you can still preview imperfect output.
• Quick layout previews are kept separate from your saved diagram, so fast drafts do not overwrite your final layout.
• Generation now restores every organisation mentioned in the conversation or validation feedback so retries stop dropping participants from multi-company diagrams.

Fixes

• Duplicate BPMN nodes, edges, and lane mismatches are cleaned up before export so downloads and validators stop breaking on duplicate IDs.
• Generation failures surface targeted tips (cross-pool flows, placeholder nodes, missing start/end events) to show what to fix before retrying.

Highlights

• Diagram generation now self-corrects common validation issues, reducing failed runs and getting you to a usable draft faster.
• Exports are checked for BPMN compatibility before you download, so they open cleanly in modeling tools without manual fixes.
• Nested subprocess steps are saved alongside the main diagram, making it easier to drill into detailed flows without losing context.
• Diagram previews load faster with automatic layout so you can review a draft immediately after generation.
• Release notes are available at /release-notes (linked in the footer) so teams can track what changed in production at a glance.
• The editor warns before you close unsaved work and restores quality signals (ratings and guidance) while generating diagrams.
• Document uploads and chat-based extraction are more resilient, keeping long-running conversations intact even with large files.

Fixes

• Plan selection no longer gets stuck after a failed upgrade attempt, reducing friction when switching tiers.
• Message rating controls and starter chat guidance are back, making it easier to give feedback and start the right way.
• File uploads fail fast with clearer validation so problematic documents don’t derail an ongoing chat.