What's New
What's new at devlin.ai
New features, improvements, and fixes as they ship.
November 2026
- New
devlin finds documents already in your library
Ask devlin now searches your workspace library when you ask about a process, policy or procedure, and names the document it found. You no longer need to attach a file you already uploaded.
- Improved
See every xAPI statement devlin.ai sends
The docs now list each xAPI statement, when it is sent, and the full default set for one attempt, so your data team can build reports without guessing.
- New
Sub-workspaces available on Enterprise
Enterprise workspaces can now hold sub-workspaces, each with its own content, people, and results. Sub-workspaces share the plan, seats, and credit pool of the parent workspace. The parent workspace determines how each sub-workspace can spend credits, it can send or mirror content to them, and it shows usage by sub-workspace.
October 2026
- New
LMS packages now send detailed xAPI results
Each attempt reports rubric criteria, mode and the learner's LMS identity, and Team and Enterprise workspaces can also send statements to their own LRS.
- Improved
Export skill results by date and learner group
The Skills tab in Brain now filters skill scores by learner identity fields such as location or role group. Export CSV downloads every skill's score for a date range, with no learner identifiers, so the file can be shared.
Coach variables based on evaluator deductions are set correctly
Coaches now see whether each deduction in a linked evaluator result was applied, so variables that depend on a deduction no longer turn on when it wasn't.
- Improved
Voice sessions wrap up gracefully at the time limit
Near the time limit the character starts closing the conversation, and a line still playing when time runs out now finishes before the session ends. Simulations devlin builds also get a little extra time beyond their seat time.
- New
Speak to devlin instead of typing
Select the microphone in the Ask devlin message box to dictate. Your words appear in the box to edit before you send.
- Improved
Simpler start screens
Learners see how they'll be scored in one line and can open a short list of the criteria. The start screen also grows to fit larger screens.
- Improved
Journey results show simulation feedback
The results page shows each simulation's feedback, and a learner who does not pass sees a fitting heading instead of a celebration. Restarting from the results no longer asks for confirmation.
- Improved
Credits and open editors update as devlin works
The credits balance and an open journey, simulation, coach or evaluator editor show devlin's changes without a page reload. Edits you have not saved are never replaced without asking.
- Improved
devlin keeps your brand and puts your character on the stage
When you share a style guide, devlin saves its colors for every build, and a journey built around a simulation shows its character on the stage.
Coaches debrief the right attempt and quote the right speaker
A voice coach opened right after an assessment now discusses that result, not an earlier one. Coaches also keep straight which lines the learner said and which the character said, and text replies no longer show extra blank lines.
- New
Ask devlin builds simulations and journeys for you
On Core, Team and Enterprise plans, describe the problem you want to solve and devlin drafts the simulation, coach or journey for you to review. Nothing goes live until you approve its changes.
- New
Journeys
Combine lessons, simulations, coaches and knowledge checks into one branded learning experience with a scorecard. Share it with a link or add it to your LMS.
- Improved
Delete test attempts from evaluator results and learner history
Owners and admins can delete a single evaluator result, or any attempt from Results > Learners, for example a colleague's test run. Deleting a simulation attempt also removes its wrap-up evaluator result.
- Improved
Class debriefs now cover every activity on a results link
The AI debrief on a results link now reads all of the link's simulations and evaluators together and shows how learners changed across attempts. It also tells the trainer exactly what it read.
- New
Merge learner records
Owners and admins can merge learners whose name or ID was typed differently on different attempts into one learner.
- Improved
Results links now include evaluators
Trainers can now see evaluator submissions on a results link alongside simulation attempts, including the combined verdict for wrap-up evaluators. Filter the feed by activity and download what you see as one CSV.
- New
Track skills and business goals across your simulations
Link evaluation criteria to your workspace's skills and sub-skills, import your own competency model, and see each skill's blended score over time. Tie skills to the business goals that they impact and track movement over time.
- Improved
Brain has its own place in the top bar
Knowledge, audiences, characters and brand settings now live together in Brain, next to Studio.
- New
Cantonese, Traditional Chinese and 15 more simulation languages
Simulations run in 33 languages, including Cantonese (Hong Kong) and Chinese (Mandarin, Traditional). Cantonese characters chat in everyday Cantonese, and a few languages, Cantonese among them, run in text only.
Coaches now see evaluator results that finish after the coach opens
A coach that opened while an evaluator was still scoring now picks up the result before the learner's next message. Voice coach transcripts also no longer show stray brackets or repeated lines.
- New
Build from meeting recordings
Add an audio or video recording of a meeting as a source. It is transcribed with speakers and times, and drafts are built from what was said.
- New
Transcript files work as source materials
Add a .vtt or .srt transcript exported from your meeting tool as a source.
PDFs with hidden characters no longer fail to upload
Some PDFs carried invisible characters that stopped them from being processed as source materials. These files now upload and process normally.
- New
Share live results with trainers
Create a results link from any simulation's Results page. Trainers and managers can view attempts, download the CSV, and get an AI class debrief without signing in.
- New
Use Rise 360 and Storyline courses as source material
Add the .zip you published from Rise 360 or Storyline and devlin.ai reads the lessons and slides in order, so new practice can be built from courses you already have.
- New
Documentation at docs.devlin.ai
Guides for every part of devlin.ai are now in one searchable place, and each page is also available as markdown for AI tools.
- New
Choose whether voice sessions can fall back to text
A new option in the voice settings controls whether learners can continue in text when the microphone is blocked or a voice session drops. It is on by default, so nothing changes unless you turn it off.
- New
Export or delete one learner's data
Owners and admins can download everything recorded for a learner, or permanently delete it, from Results, Learners.
- Improved
Version history shows when a simulation was unpublished
Taking a published simulation or coach offline is now recorded in its version history.
- New
Leave an organization yourself
Members and admins can leave an organization from Settings, Workspaces. Owners transfer ownership or delete the organization instead.
Weekly digest unsubscribe affects only you
The unsubscribe link in the weekly digest turns the digest off for you alone. It no longer changes what your teammates receive.
- New
Light theme
You can now switch devlin.ai between Dark, Light, and System from the account menu. Your choice is saved to your account and follows you to the docs and marketing pages.
- Improved
Review comments now appear in the editor
Comments left on a review link show up in the Share for review card, and stay there after the link expires or is revoked.
- New
Evaluators are available on every plan
Score work your learners submit, like call notes or a completed assessment, against your own rubric. Use an evaluator on its own or link it to a simulation for one combined result.
- New
Read coach session transcripts in Results
Each coach now lists its learner sessions in Results, and you can open any session to read the full conversation.
- Improved
Easier to read simulation results
Criteria reasoning, quoted evidence, and learner responses in results are larger and higher contrast.
- Improved
Keyboard and screen reader access across the app
Every page was checked against WCAG 2.1 AA. Labels, contrast, and keyboard focus were fixed throughout the builder and learner pages.
- New
Give a simulation's evaluation its own reference documents
You can now attach reference documents for scoring directly in a simulation's Evaluation tab. They are kept separate from the documents the character uses in the Knowledge tab.
September 2026
- New
Show learners what they'll be scored on
Turn on criteria before the simulation, during it, or both. Learners see each criterion and its points on the start screen and from a Criteria link in the chat, in their language when you've added languages.
- New
Evaluations can check answers against your documents
Attach policies or guides to an evaluation and optionally let the AI evaluator use them to check accuracy. You can also draft assessment criteria straight from those materials.
Attempt exports show each attempt's language again
The language column in attempt exports and the Metrics tab now shows the language each learner used.
Criterion points in attempt exports open correctly in Excel
Points now read as "8 of 15", so Excel no longer turns them into dates.
- New
Build from your own documents and links
Add PDFs, Word, PowerPoint, Excel files, or web links when you create a simulation or coach. Keep them in your library for reuse, or use them just for the build.
- New
Learners can see how they did on each criterion
Turn on the criteria breakdown and learners see each criterion, their rating, and a short reason after every attempt. It appears in the learner's language when you've added languages to the simulation.
- New
Limit how many attempts learners get
Set a maximum number of attempts per simulation, counted per browser or per learner. Learners see which attempt they're on, and Try Again is hidden on the last one.
- Improved
Results and Publish now include hosted projects
Results now shows views for your hosted projects alongside simulations and coaches, and Publish lists them with quick actions to copy, open, or manage each one. You can also drag the sections on both pages into the order you want.
- New
Learners can review their simulation transcripts during coaching
Coaches can now show a Simulation transcript button, so learners can reread their completed conversation while they're coached, even when the simulation is on a different slide. Turn it on for each linked simulation in the coach's Linked Simulations tab.
Coach replies no longer start with blank lines
Coach messages in text mode now show without extra empty space above or below the reply.
Large documents now upload to knowledge bases
Knowledge base documents over 4.5 MB now upload successfully, up to your plan's limit of 75 MB on Core or 150 MB on Team and Enterprise.
- New
AI tools can now host your pages, connect your domain, and give you the right link
Assistants connected through the MCP server now give you a test link for every new simulation and the share link after publishing or editing a published simulation. They can also publish a web page or site on devlin.ai hosting, rename its address, and walk a workspace owner or admin through connecting a custom domain.
- New
AI learner summaries on the Learners page
Generate a short handoff summary of any learner's attempts across all simulations and evaluators, then copy it into your notes. The credit cost is shown before you generate.
- New
Learner identity is now available to every workspace
Workspace owners and admins can turn on learner identity collection under Settings, Data & privacy. Once on, each simulation chooses whether to ask learners for an identifier and supporting fields on its start screen.
- Improved
Start screen and Variables have their own sections in the simulation editor
The start screen now has a live preview and its own place in the Design rail, and simulation variables moved out of the Publish page into Design, Variables.
- Improved
Coach voice sessions can now be resumed after a dropped connection
If a voice conversation with a coach loses its connection, learners can pick up right where they left off instead of starting over. The coach remembers the conversation so far.
- New
Start screens can now collect role, location, and other learner details
Workspaces can define up to five identity fields, choose which ones each simulation asks for, and filter results and CSV exports by them.
- New
Serve one simulation in several languages
Add languages to a simulation and send each course version to the same link with a language code. Learners get the opening line, the widget, the character, and their feedback report in that language, in text and voice.
Coaches start a new session after a course restart
When a learner restarts a course and resubmits to the evaluator or retakes a simulation, the linked coach now opens a fresh coaching session instead of staying locked as ended.
- New
Download receipts and invoices from Credits & Billing
Every payment on your workspace, including credit top-ups, now has a receipt link on the Credits & Billing page. Subscription payments also include a downloadable invoice PDF.
August 2026
- New
Coaches can now update Storyline variables and end sessions
Define ending conditions and custom variables for coaches, just like simulations. Coaches report completion, turn count, timing, and your custom variables back to your course, and the coach editor's Test tab shows each variable change live.
- New
See when your character is speaking
A new variable is set to true while the AI character speaks in voice mode, for both simulations and coaches, so your course can react in real time.
- Improved
Voice mode now follows your theme everywhere
Coach voice screens are fully customizable in the visual editor, and the mode-switch buttons and error screens now use your colors with accessible contrast built in.
Publishing is more reliable when content screening has issues
If the content screening step hits a temporary problem while publishing a simulation, coach, or evaluator, we now automatically retry with a backup model instead of asking you to try again later.
- New
Require approval before AI tools change published content
Workspace owners and admins can now require that any change an AI tool makes to a published simulation, coach, or evaluator over MCP is reviewed and approved in devlin.ai first. Every request shows the exact before and after, and nothing is applied until someone approves it.
- New
Connect ChatGPT and Claude directly to your workspace
Workspaces with MCP server access can now add devlin.ai as a custom connector in ChatGPT Enterprise and claude.ai. Sign in with your normal devlin.ai account, pick a workspace, and you're connected. No tokens to copy or rotate.
- New
See every attempt by a learner in one place
Results now has a Learners view that lists each learner's simulation attempts and evaluator results together, with duration, wrap-up time, criteria, and feedback per attempt. Two CSV exports give you one row per attempt or one row per criterion, ready to paste into a training tracker.
- Improved
Duplicated simulations and coaches get their own voice agent
Copies are ready for voice testing right away, and a copy can no longer interfere with the original's voice sessions.
- Improved
Voice connection errors now say why
When a voice session drops, the connection lost screen and the attempt record include the voice provider's reason, so problems can be diagnosed without a support round trip.
- New
Warning when a voice character is flagged by the voice provider's safety check
The warning appears in the simulation's voice settings, so you know before learners run into a session that will not connect.
- Improved
Low credit alerts arrive immediately
You now get an email the moment your balance drops below a threshold, and an alert if a session is declined for insufficient credits. Alert thresholds are configurable under Settings, then Credits.
Voice simulation feedback now scrolls
Long feedback shown at the end of a voice simulation can now be scrolled. Previously, feedback that did not fit on screen was cut off.
- New
Custom rating scales for simulation criteria
Choose how many credit levels each evaluation criterion has, what each level is worth, and how it is labeled on results. Partial credit no longer has to mean 50%.
- New
Private scoring guidance for rubric criteria and deductions
Add judge-only guidance to any criterion or deduction to calibrate how the AI scores it. Learners and results pages only ever see the plain rubric description.
- New
Evaluate on session length
The evaluator now knows exactly how long each session lasted, so your criteria and deductions can reference call or chat duration, like deducting points when a call runs long.
- New
Get notified before a simulation hits its daily limit
Workspace owners and admins now get an in-app and email alert when a simulation reaches 80% of its daily credit limit, with a link to adjust it.
- Improved
Set daily spending caps on your simulations and coaches
Workspace owners and admins can now set a daily credit cap on each simulation and coach from its Publishing tab, plus a workspace-wide default for simulations under Credits and Billing settings.
- Improved
Clearer message when a simulation hits its daily limit
Learners now see a daily usage limit message instead of a generic voice unavailable screen, so you can tell a spending limit from a technical issue.
- New
Character pronouns
Give your character pronouns and they'll describe themselves accordingly.
- New
Sim_Timer course variable
Sims can report elapsed time to your course every second, so your own on-screen timers stay in sync. The clock starts at the learner's first message in text sims, or at session start in voice sims.
- Improved
Delete test attempts from the Metrics tab
Workspace owners and admins can remove individual attempts, such as their own test runs, and every metric on the page updates immediately.
- New
Start and end voice sessions from your own course buttons
Wire course buttons to startVoiceSession() and endVoiceSession() in Storyline so learners never have to find the widget's controls.
- New
Let learners choose text or voice mode
Simulations and coaches can now offer learners a choice of text or voice before a session starts. New toggles in the editor control which modes each link offers.
- Improved
Test voice mode on drafts
The Test stage in the simulation and coach editors now has a Text/Voice picker, so you can try voice conversations before publishing.
Voice sessions now explain when they can't start
If a voice session fails to start, learners see a clear message with a retry option instead of being switched to text with no explanation.
- New
Let search engines find your hosted projects
Hosted projects now have a search engine toggle. Projects stay hidden from search results by default, and you can turn on indexing for portfolios and public work.
- New
Get a new link for a hosted project
If a hosted project link ends up somewhere you did not intend, you can now generate a new link. The old link stops working immediately, everywhere it was shared.
July 2026
- New
See variable values live while testing a simulation
The Test view now lists every course variable your simulation can set, both automatic and custom, with values that update in real time as you chat. You can check your variable logic without embedding the sim in Storyline first.
- New
New Sim_SessionStart course variable
Your course can now react the moment a learner actually begins a simulation. Sim_SessionStart is set to true on the first message (or when a voice session connects) and resets when the simulation restarts.
- New
Choose how long review links last
Pick an expiry from 1 day to 90 days when creating a review link. Links still default to 2 weeks.
Background color changes save cleanly
Typing a hex code for a hosted project's background no longer triggers repeated error pop-ups. The color saves once you finish typing.
Social preview images display in the dashboard
The social preview card now shows your uploaded image reliably inside the platform.
- Improved
Editors open each section at the top
Switching sections in the simulation and coach editors now brings you to the top of the page.
- New
Try devlin.ai right from the homepage
Describe a scenario on the homepage and play a live simulation instantly, no account needed. Sign up to keep the simulation you built.
- New
Voice mode for coaches
Coaches can now talk learners through debriefs and direct conversations, drawing on their linked knowledge just like in text chat. During a live simulation, coaching stays in text so your coach never talks over the sim.
- New
Character portraits with live expressions
Give your simulation's character a face. Pick from the built-in portrait library and the character's expression changes as their mood shifts during the conversation. You control the mood variable, the starting mood, and which expressions are available.
- New
Customize how your hosted project links look when shared
On the Starter plan and above, set a custom title, description, and preview image for each hosted project. Shared links now unfurl with your project name by default.
- New
Connect Power BI to your workspace analytics
Enterprise workspaces can now generate read-only API keys and pull simulation rollups, daily activity, and learner attempt data into Power BI. Copy-paste Power Query snippets are included in the new Integrations tab.
- New
Protect your account with two-factor authentication
You can now add an authenticator app to your devlin.ai account in Settings. Once it's on, signing in requires a one-time code, keeping your simulations and learner data safe even if your email is compromised.
- New
Upload project folders to hosting
You can now upload your published project as a folder, no zipping required. Choose Upload folder on the Hosting page, drag and drop the folder onto it, or use either option when uploading a new version.
Cleaner focus styling in text simulations
The chat input no longer shows a stray blue glow around the message box while you or your learners type. The intended focus ring still appears.
- New
Share your work for stakeholder review
Generate a review link for any simulation and collect threaded comments right beside a live preview. Reviewers do not need an account, your workspace gets notified when feedback lands, and you can resolve threads as you address them.
- Improved
A simpler, more focused devlin.ai
The app now centers on Studio, Results, and Publish. Everything you build lives in Studio, so there is less sifting through menus.
- New
See your published projects in one place
The Publish page now opens with an overview of everything you have published, flags drafts with unpublished changes, and lets you publish updates right from the list.
Evaluation feedback no longer shows internal turn numbers
Learner feedback now describes moments naturally, like 'early in the conversation', instead of internal labels such as T2 or T6. Perfect performances also no longer receive made-up improvement suggestions.
- New
Starter plan at $9/month
A lower-cost plan for solo designers building a portfolio. Includes 5,000 credits each month, or 60,000 up front on the $72/year annual option.
- New
Set a data retention window for learner transcripts
Organization workspaces on the Team and Enterprise plans can choose how long learner conversations and evaluation feedback are kept. Older data is deleted automatically every night while your metrics stay intact.
- New
Delete your account yourself
Permanently delete your account and all of its data from the account page, no support email required.
- Improved
More consistent and accurate learner evaluations
We hardened our evaluation engine and now regression test it against a library of graded practice sessions. Learner performance is scored consistently and accurately, and every piece of cited evidence is verified against the actual transcript.
- New
One-way simulation variables
True/False variables can now be locked to change in one direction only. Once a milestone like Door_Opened switches to true, it can never flip back for the rest of the session.
- New
Turn counter variable
Simulations can now send the current turn number to your course as it happens. Use it to warn learners when they are running low on turns, or to branch course logic by progress.
- New
Workspace analytics
See results across your whole workspace in one place. Sessions, completion rate, pass rate, average score, and credit usage, with per-simulation and per-coach breakdowns and CSV export.
- Improved
Hosted projects load faster
Files in hosted projects are now compressed, so published courses load noticeably faster for learners, especially on slower connections.
- New
Download your original files
You can now download the original files for any hosted project from its detail page, as a single file or a zip.
- New
Serve hosted projects on your own domain
Bind each hosted project to one of your verified custom domains with its own path. You can also choose which project loads when someone visits the bare domain.
- New
Enterprise single sign-on (SAML SSO)
Organizations on the Enterprise plan can sign in through their identity provider, such as Okta or Microsoft Entra ID. Admins can require SSO for their team, and teammates are added to the organization automatically on first sign-in.
June 2026
- New
Host your Claude Design & Storyline 360 courses on devlin.ai
Upload a Claude Design course exported as HTML, or a Storyline 360 or Rise package, and publish it at a shareable link with no separate web host required. Hosted projects work as standalone pages or embedded in your LMS.
- Improved
A cleaner simulation metrics page with full attempt history
The metrics tab now shows your complete attempt history, including attempts from before credit tracking began. The date range filter scopes the whole page, and the layout is easier to scan.
- New
See exactly where your credits go
Simulation and coach metrics now show credits per session and per completion with a date filter, an attempts list, and per turn costs in transcripts. You can expand any day on the credits page to see the split by simulation, coach, and evaluation, and export attempts or transcripts to CSV.
- New
Wrap your Claude Design courses as SCORM packages
Use the new SCORM Wrap tool in your dashboard to turn a Claude Design course, exported as a single HTML file, into a SCORM package that reports completion and scores to your LMS. Wrapping happens right in your browser, so your course never leaves your device.
- New
Introducing Knowledge Bases
Upload reference documents and group them into reusable knowledge bases, then attach them to any simulation or coach. The AI grounds its responses in your material and optionally cites the source, so answers stay accurate to your content.
- Improved
Coaches now align their guidance with how a simulation is graded
During live practice, a coach can steer learners toward what the simulation actually scores, and in the debrief it can walk through performance criterion by criterion. You can also turn off live criteria sharing per simulation or make a coach debrief-only.
- Improved
Coaches can be tuned separately for live help and debriefs
You can now set different guidance styles for when a learner is mid-simulation versus reviewing a finished attempt. Coaches also stay brief automatically during live voice simulations so they don't stall the conversation.
- Improved
A refreshed, easier to navigate builder
The dashboard got a redesign. A persistent sidebar puts every page one click away, simulations can show as a list or tile grid, and the editor pages are roomier and make better use of large screens.
- New
Folders for organizing simulations and coaches
Create folders right from the dashboard, then drag a card onto a folder to file it. You can also select several items with checkboxes and move, duplicate, or delete them together.
- New
Command palette for quick navigation
Press Cmd+K (Ctrl+K on Windows) anywhere in the app to search your simulations and coaches by name or jump straight to any page.
- New
Simulation version history
The editor now keeps a history of every save and publish. Open Version history to restore an earlier version, and restoring is itself recorded so you can always step back again.
- New
Export metrics to CSV
The Metrics tab now has an Export CSV button so you can analyze learner results in Excel, Google Sheets, or your BI tool.
- New
Voice mode preview in the Visual Editor
See how voice simulations will look while you style them. The Visual Editor shows a voice preview alongside the text preview, with toggles for in call, AI speaking, and end of simulation states.
- New
Light and dark theme presets in the Visual Editor
Apply a polished light or dark starting theme with one click, then fine-tune individual colors from there.
Coaches now start fresh when a course is restarted
Real-time coaching now responds to your current conversation instead of referencing an earlier attempt after a course is refreshed. Coaches still remember the conversation when learners move between slides.
- Improved
Accessibility upgrade across the learner experience
The voice, text, and coach simulation widgets are now built to WCAG 2.0 AA. Learners using screen readers, keyboard navigation, or reduced motion get a first-class experience.
- New
Contrast checks in the appearance editor
When you customize a simulation's colors, the appearance editor now checks them against WCAG AA contrast standards and flags combinations that fall short. Most can be fixed with a single click.
Voice previews play again in the simulation editor
Clicking a voice in the picker now plays its sample again. The preview audio had stopped loading after a change on our voice provider's side.
Voice simulations now sit flush inside Storyline and Rise courses
When embedded in a course, a voice simulation now fills its frame cleanly instead of showing a floating card with misaligned borders. The standalone voice link still opens as a centered card.
- New
Introducing coaches, AI mentors that debrief your learners
Build an AI coach that reviews how a learner did across your linked simulations and walks them through what to improve. Coaches give real time guidance during voice simulations and a personalized debrief afterward, and you can embed or export them to your LMS like any simulation.
- New
Export voice simulations to your LMS
You can now download a SCORM or xAPI package that launches a voice simulation, not just text chat. Upload it to your LMS the same way, and pass/fail and score report back.
May 2026
Voice simulations can now be retried
Voice simulations now show a Try Again button when they end, just like text simulations, so learners can start a fresh attempt. Turn it on with the Try Again setting in the sim editor.
Storyline variables in voice simulations no longer flicker mid-conversation
In voice simulations, the variables sent to Storyline could change several times while the learner was still speaking, briefly flickering or showing the wrong value. Each variable now updates once when the character responds.
Transcripts and scores now reflect the real conversation
Pauses, retries, or reconnects could record the same message more than once or split a sentence into pieces. That cluttered the transcript and sometimes lowered a learner's score. Each turn is now saved once, and voice evaluations use the finalized conversation, so scores and transcripts match what was actually said.
- New
Multilingual simulations
Simulations now support 16 languages including Spanish, French, German, Japanese, Arabic, and more. Set the language in the sim editor, and all UI elements, conversation, and evaluation feedback render in that language. Text-mode embeds also accept a ?lang= query parameter for language overrides.
The AI wizard occasionally failed to generate a simulation
This could happen when designers didn't give the AI enough information to work with. It now handles those cases reliably instead of erroring out.
- New
Voice mode is now available to all users
Voice mode is now live for every account. Turn it on for any simulation with the toggle at the top of the Simulation tab.
- New
Change your account email
You can now update the email address on your account from account settings.
Evaluations no longer penalize learners for not replying to the final message
Previously the AI could lower a learner's score for not responding to its closing message, even though the simulation had already ended. Learners are now scored only on exchanges where they had a real chance to respond.
- Improved
Voice mode now costs 20% less
Reduced from 150 to 120 credits per minute.
Scores now display as percentages
Now shown as a percentage (e.g. 72%) in the Metrics tab, the learner's end-of-session feedback, and the Storyline Sim_Score variable, instead of raw points.
The Credits page no longer double-counts voice sessions
Usage history now matches the credits actually deducted from your balance.
Voice credits are refunded if a session ends unexpectedly
If a session ends without closing properly, such as a crash or lost connection, reserved credits are returned automatically.
- Improved
Longer conversations cost less
Credit usage now benefits from prompt caching, so longer simulations cost less per turn.
- Improved
All data processing is now US-only
Removed Google Analytics, the embedded YouTube video, and the Google Fonts link, so every request from designers and learners stays on US-based infrastructure.
- Improved
Voice privacy: shorter retention, no audio storage
Voice transcripts are kept for only 30 days, and session audio is never stored.
April 2026
- New
Voice mode now works inside Storyline embeds
Learners can hold a spoken conversation right inside a Storyline course, not only in the standalone chat.
Voice mode reliability fixes
Voice sessions now correctly let the learner speak first when set up that way, and if a learner interrupts the AI, that part of the conversation is no longer lost from the transcript.
The editor shows a saving spinner instead of a confusing save/discard prompt
While your changes are saving, the editor now shows a spinner instead of briefly asking whether to save or discard.
Fixed a layout glitch in the editor preview when the branded outline is off
When a designer turned off the branded outline, the simulation preview in the editor showed the chat bubbles and input side by side instead of stacked. They now stack correctly.
- New
Team workspaces
Create an organization, invite teammates by email, and share a workspace with pooled credits. Owners and admins can manage members, roles, and invitations.
- New
Voice mode (early beta)
Run your simulations as a spoken conversation instead of typing.
- New
Simulations can report SCORM data to LMS from within Storyline
When a simulation is embedded in a Storyline course, it can now send the learner's score, pass/fail, and completion straight to your LMS through Storyline's existing SCORM connection.
- New
Core plan launched
The Core plan is now available, unlocking the full simulation builder.
- New
Branded simulation outline
Simulations now include a default gradient outline in devlin.ai's brand colors, framing the conversation. You can turn it off per simulation.
- New
devlin.ai public beta launched 🎉
devlin.ai opened to the public, so anyone can sign up and start building simulations.