Hyper 1.12.0

Goal for this release

The last release brought cloud browser use, more robust Hypercards, personalized suggestions, and longer-running tasks.

This release brings our new branding across the desktop app and website, introduces paid plans, adds four connectors, and makes conversations faster to open and easier to navigate. It also improves model selection, attachments, large tool results, onboarding, and everyday reliability.

Our goals were to:

• Bring the desktop app and website into one visual style.
• Give users a free plan and paid plans for different amounts of work.
• Make usage, model access, and plan management clear.
• Expand the services Hyper can work with.
• Make long conversations faster to load and easier to navigate.
• Preserve drafts, attachments, and active work as you move through the app.
• Improve reliability across onboarding, connected tools, and older conversations.

Paid plans and a free trial

Hyper now offers Common, Rare, Epic, and Legend.

Common is free. The paid plans provide more usage, additional models, cloud computer and browser access, and higher limits for background work and automations.

Billing belongs to each member, across the workspaces they use. Joining or creating another workspace does not create a separate subscription or usage allowance. ( #1450, #1454, #1459, #1584)

Choose a plan

Local computer use remains available on every plan.

Paid plans add cloud computer and browser sessions and background agents. Higher plans allow longer cloud sessions and more concurrent sessions.

Automation limits are:

• Common: Up to five active automations, with daily schedules.
• Rare: Up to ten active automations.
• Epic: Up to 25 active automations.
• Legend: Unlimited active automations.

Paid plans also support weekly and monthly schedules.

See the full plan comparison. ( #1454, #1458, #1522, #1579, #1584)

Time to choose your plan

Existing users receive four days of Epic with no usage limit to try the plan and decide whether to upgrade.

New users receive a seven-day Epic trial after completing onboarding.

No payment card is required. When the trial ends, users who have not chosen a paid plan move to Common. Saved work remains available, while new work follows Common’s usage, model, and feature limits.

The trial removes the included-usage cap. Epic’s other feature and session limits still apply. ( #1527, #1538, #1543, #1551, #1581, #1603)

Update the app before your trial ends

Fully quit and reopen Hyper to install the update.

Older app versions that cannot show the new billing state can continue during the trial. After the trial, the server requires an updated app before accepting further billable work.

The error explains that Hyper needs to be fully quit and restarted. This does not delete the account or its saved conversations. ( #1603, #1609)

See your usage in Settings

Settings now shows:

• Your current plan.
• The percentage of included usage used.
• Your usage reset date.
• Your trial end date, when applicable.
• Links to manage or upgrade your plan.
• Extra-usage controls when available.

The app shows usage percentages rather than internal usage units or per-message costs.

On iPhone, Settings also shows your plan and usage status. Plan management remains in the desktop flow. ( #1526, #1540, #1561, #1591, #1612)

Buy extra usage when you need it

Paid plans can buy prepaid extra usage through Stripe.

Hyper uses the included allowance first, then an available extra-usage balance. The extra balance carries across monthly allowance resets and is shared across the member’s workspaces.

Purchasing extra usage adds usage to the current plan. It does not unlock models or features from a higher plan.

Payment confirmations are processed once, so receiving the same payment event again does not credit the purchase twice. ( #1528, #1547, #1550, #1554, #1571)

Usage limits apply to new work

Hyper checks plan access and available usage before starting work.

A turn that has already been accepted can finish. When the allowance and available extra usage are exhausted, further work requires a reset, an upgrade, or an eligible top-up.

Usage accounting includes the work needed to carry out a task, rather than counting only the visible final reply. Subscription renewals and plan changes update the member’s billing state through Stripe events. ( #1475, #1476, #1511, #1513, #1541, #1556, #1594)

A completely redesigned website

We rebuilt heyhyper.ai with a new visual style, illustrations, interactive scenes, and product demonstrations.

The redesign previewed in the last release is now implemented across the main website, including the home page, founders page, download flow, FAQ, and supporting pages. ( #1427, #1448, #1449, #1455, #1490)

See Hyper in action

The home page includes a TV-style opening, product demonstrations, and role-based examples of work Hyper can help complete.

The examples cover work such as engineering, sales, marketing, recruiting, and company operations.

These demonstrations illustrate product interactions. They do not execute tasks against a visitor’s connected accounts. ( #1451, #1461, #1483, #1496, #1505, #1515, #1582)

An interactive FAQ

The FAQ presents answers through a chat-style interface.

You can browse questions, filter them by typing, and open the answer you need. On smaller screens, the question list moves into a menu.

The answers are prepared product information, not responses generated by a live AI chatbot. ( #1456, #1484, #1557)

New illustrations and more consistent motion

We added new character illustrations, founder portraits, animated scenes, keyboard interactions, and optional sound treatments.

Follow-up improvements refine responsive layouts, transitions, video framing, dark and light media, and the closing sections of the site.

We also improved how the site serves its media and how its pages appear in search results. ( #1441, #1445, #1446, #1477, #1497, #1499, #1500, #1521, #1524, #1569, #1592)

A completely new look for Hyper

The desktop app now uses our comic-inspired visual style across its main surfaces.

This includes:

• Colors, borders, shadows, and icons.
• The sidebar and navigation controls.
• Settings and onboarding.
• The chat bar and model menu.
• Hypercards and tool activity.
• Links, attachments, menus, and tooltips.

The refresh applies to the existing product rather than requiring you to recreate your conversations or settings. ( #1564, #1570, #1575, #1586, #1588, #1608)

Refreshed Hypercards and tool activity

Hypercards, approval controls, tool steps, and subagent activity now share more consistent styling.

Connector logos have improved dark-mode artwork where needed. Choices, controls, and activity indicators also use the updated visual system.

These changes improve presentation; they do not introduce a new category of connector actions. ( #1482, #1601, #1614, #1615)

A redesigned chat bar

The chat bar has a flatter layout, refreshed attachment cards, and a redesigned model menu.

The model menu groups choices by provider and identifies models that require another plan.

In a conversation already bound to OpenAI or Anthropic, it shows that provider’s models and explains why the other provider is not available in the same chat. ( #1606)

Updated iPhone styling

The iPhone app receives updated color assets, chart colors, and styling in its embedded editor.

This release does not replace every native iPhone screen with the desktop redesign. ( #1564, #1570)

New model and easier model choices

GPT-6.1 Terra

We upgraded:

• GPT-5.6 Terra → GPT-6.1 Terra

Saved selections of the previous Terra model resolve to the replacement.

GPT-6.1 Sol remains the general default before plan restrictions are applied. Common users receive an allowed model rather than being left with an unavailable selection. ( #1567, #1589)

Remember your selected model

Hyper now remembers the selected model between app sessions.

The model picker also reflects the current plan. Models outside that plan show which plan includes them.

When a plan change makes a selected model unavailable, Hyper resolves the selection to an allowed model from the same provider. This keeps the conversation usable after a trial or higher-tier plan ends. ( #1523, #1589)

Choose a model for an automation

Automation editing now includes model selection.

You can choose a model available on your plan instead of relying only on an implicit background choice. Automation execution also checks the initiating member’s plan access. ( #1574, #1576)

Four new connectors

Hyper now connects to Bookface (for all our YC folks), X, Clay, and Smartlead.

Supported operations depend on the connected account’s permissions, the provider’s limits, and Hyper’s action controls.

Bookface

Connect Bookface to work with the YC founder community.

Hyper can search companies, founders, investors, deals, forum posts, and jobs. It can also read information available to your account, including office hours, batch schedules, fundraising records, hiring information, and YC documents.

Supported actions include:

• Updating company and founder profiles.
• Posting company news.
• Claiming or redeeming deals.
• Booking office hours.
• Managing supported fundraising and sales records.
• Managing jobs and candidates.
• Responding to events and founder dinners.

Hyper indexes your own company and founder profiles. It does not copy the whole Bookface community into its search index.

Bookface changes and quota-spending contact reveals require approval. ( #1474)

X

Connect X to search posts and work with your account.

Hyper can read your timeline, mentions, bookmarks, lists, followers, and direct messages, as well as supported public account, post, and analytics information.

Supported actions include posting, replying, quoting, reposting, liking, bookmarking, following, and sending direct messages.

For example:

Find recent posts discussing our launch and summarize the feedback.

Draft a reply to this post for me to review.

X is read on demand rather than copied into a background index. Uploading images or videos from your computer through this connector is not included. ( #1514, #1531)

Clay

Connect Clay to find people and companies and enrich the results with additional information.

Hyper can search public people and company data, read enrichment results, inspect available credits, and work with supported Clay Audiences information.

It can also run functions your team has configured in Clay.

Enrichments and function runs can spend Clay credits. This connector does not read or edit Clay tables.

Clay is read on demand rather than indexed in the background. ( #1562)

Smartlead

Connect Smartlead to review outreach performance and work with campaigns, leads, replies, and mailboxes.

Hyper can:

• Read campaigns, sequences, schedules, and settings.
• Review replies, sent mail, and lead message history.
• Inspect campaign and mailbox analytics.
• Create and update campaigns and sequences.
• Add and manage leads.
• Reply to or forward lead emails.
• Start or pause campaigns.
• Search Smart Prospect and reveal available email addresses.

Starting a campaign sends real email according to its configuration. Smart Prospect email reveals can spend provider credits.

Adding mailboxes and managing Smartlead clients are not included. Smartlead is read on demand rather than copied into a background index. ( #1565)

Faster loading and easier conversation navigation

Open long conversations faster

Hyper no longer needs to load every message before opening a long conversation.

Desktop and iPhone load a recent window of messages and retrieve older history as needed. Loading an older page preserves the reading position instead of moving the conversation unexpectedly.

This reduces the amount of history that must be fetched and displayed before you can resume work. ( #1595, #1598)

Jump to any turn on desktop

The desktop navigation rail can list every turn independently of whether its messages are loaded.

You can:

• Hover over a marker to preview its title.
• Select a marker to jump to that turn.
• Scroll over the rail to move through turns.
• Return to the latest messages.

Hyper loads the relevant messages when you jump. You can then continue scrolling in either direction.

New replies leave you in place while you read earlier history. ( #1611)

Smoother streaming and scrolling

We reduced the work required to update the conversation while a reply streams.

Desktop also avoids keeping every distant turn fully rendered at once, while preserving its space in the conversation.

Together, these changes reduce repeated rendering work and improve responsiveness in long chats. ( #1444, #1599)

Comments and cards stay attached to the right content

Messages now retain stable positions as history loads in separate pages.

Comments remain attached to their original messages instead of depending on a message’s temporary position in the loaded transcript.

Card revisions also update the relevant content while you are reading an older part of a conversation. ( #1597, #1611)

Better handling of large results

Large text and JSON results from tools can now be kept as conversation files.

Hyper receives a preview and a reference to the full result. It can then inspect the relevant parts as it continues the task.

This keeps the original result available without placing all of it into the immediate conversation context.

The change applies to tool output. It does not remove the connected service’s pagination, access, or file-size limits. ( #1613)

Easier attachments and more reliable drafts

Add files from the system file picker

The desktop chat bar now opens the system file picker to add attachments.

Selected files are checked against the supported file types and the remaining room in the draft. Files that cannot be read or do not fit produce a clear refusal rather than silently disappearing.

The attachment remains associated with the draft that opened the picker, even if you move elsewhere while selecting files. ( #1606)

Keep attachments safe while sending

Adding files and sending a draft are coordinated so that a file arriving during a send is not lost.

Each draft handles its own send state. Sending a message in one chat does not unnecessarily block another chat’s draft.

These changes also account for files whose size or contents change while they are being read. ( #1606)

More reliable active work

Queued messages continue on the live stream

Eligible queued turns can continue through the existing live connection.

The server now records input identity and delivery more consistently, helping distinguish messages waiting to run from instructions directed at an active task.

This improves continuity when several messages are queued or when a request is retried. ( #1463, #1472)

Steer a task after reopening the app

Reopening a conversation now uses the server’s active-run information more reliably.

You can continue giving instructions to work that is still running instead of the client treating the conversation as idle simply because it was reopened. ( #1516)

Send feedback directly from the app

The desktop header now includes a feedback control.

Open it, write your feedback, and send it directly to the founders without leaving Hyper.

The submission includes account and app information that helps identify the problem, such as the workspace when available, app version, operating system, and an available diagnostic-session reference.

The same feedback submission is not sent twice when a completed request is retried. ( #1568)

Onboarding and welcome improvements

More reliable Gmail setup

We fixed a timing problem that could cause the getting-started Gmail card to appear before the connection was ready.

Attaching Google sign-in permissions also no longer waits for an inbox scan as part of that step.

These changes improve setup timing without turning the welcome conversation into a recurring inbox scan. ( #1432, #1435)

Better selection of emails waiting for a reply

The getting-started flow now filters out tool-generated reminder threads when choosing an email that appears to need your attention.

This reduces cases where an automated reminder is presented as a work conversation waiting on you. ( #1530)

A welcome email after signup

New accounts receive a welcome email with a short introduction, guidance on connecting tools, and a link to book a call with the founders.

Delivery is queued and retried when needed. Repeated delivery attempts use the same send identity to avoid duplicate welcome emails. ( #1533)

Privacy and diagnostic recording changes

Updated recording defaults and controls

Diagnostic recordings can include conversation text and images.

New installations now default to including this content. Existing locally saved recording preferences continue to be read.

The recording-sharing controls introduced in the previous release are no longer shown in onboarding or Settings. The temporary onboarding screen added during the patch cycle is also no longer part of the final flow. ( #1433, #1492, #1493)

Privacy information is available from Account

Account settings now provide the Privacy Policy link and a contact for privacy questions or requests.

We also corrected affected stored recording preferences and improved the styling of masked diagnostic recordings.

These changes do not introduce a new self-service recording toggle. ( #1429, #1506, #1519, #1602)

Performance and reliability fixes

More reliable Granola synchronization

Granola can return a separate section describing duplicate recordings after its meeting list.

Hyper now distinguishes that section from the actual meetings to synchronize. Duplicate-recording references are no longer parsed as incomplete meeting records.

This fixes synchronization handling; it does not delete recordings from Granola. ( #1558)

Embedded videos stay inside Hypercards

Supported video players now register their embedded pages with the desktop app before loading.

This prevents the app from sending the player to an external browser instead of showing it inside the card.

An unchanged video address also no longer reloads simply because another chat update arrives. ( #1512)

Older conversations remain readable

We improved conversion of conversations stored in the earlier message format.

The conversion preserves usable conversation history in the current format. Follow-up cleanup removes the obsolete storage path after that migration.

This is a compatibility change for existing conversations, not a request to delete their contents. ( #1572, #1578)

More consistent connector requests and approvals

Specific connector routes now take precedence over broader route patterns.

Approval details also use a stable field order, and saved tool data has more consistent serialization.

These changes reduce avoidable differences between equivalent requests and make action details easier to review. ( #1590, #1593, #1610)

Fewer unnecessary sign-outs

A generic refusal from the authentication service no longer automatically ends the session.

Hyper distinguishes these failures from conditions that require signing in again. ( #1473)

Several improvements in these notes already shipped in 1.11.1–1.11.4, including Gmail onboarding fixes, diagnostic-recording changes, and older-conversation compatibility. They are included for anyone upgrading from 1.11.0.

Other additions and improvements

• Clearer dark-mode artwork: Improved connector-logo contrast and corrected the X icon on iPhone. ( #1482, #1531, #1601)
• Updated billing pages: Refined checkout and plan-selection layouts to match the new branding. ( #1560, #1585)
• Updated terms: Revised the published fee and refund language alongside the pricing work. ( #1525)
• More consistent controls: Refined keyboard hints, floating menus, tooltips, and dark-mode borders throughout the desktop interface. ( #1588, #1606, #1614, #1615)

Engineering and release improvements

More complete usage accounting

We added durable usage recording, monthly allowance periods, prepaid-balance accounting, and subscription-event handling.

Usage writes can be retried without applying the same charge twice. Historical cost records remain separate from the customer-facing usage percentage.

We also updated model cost accounting and added monitoring for plan usage and limits. ( #1475, #1476, #1511, #1529, #1580, #1600)

Separate billing test and live environments

Local and nightly environments use Stripe test credentials. Stable uses live credentials.

Development billing controls allow internal testing of plan, trial, and usage states without exposing those controls in stable.

The release includes the implemented payment and enforcement paths, rather than only the earlier pricing mockups. ( #1537, #1542, #1549, #1555, #1594)

Faster and more focused checks

• TypeScript checks: Run more work in parallel and reuse generated bindings.
• Rust tests: Reuse built test binaries rather than rebuilding them unnecessarily.
• Database types: Improved registry access and limited regeneration checks to changes that affect their inputs.
• Protocol compatibility: Main-branch checks compare against the latest nightly baseline.
• Stable tests: Removed dependencies on map iteration order.

These changes reduce repeated CI work while retaining the relevant validation steps. ( #1510, #1539, #1548, #1559, #1563, #1605, #1607)

Clearer operational monitoring

Failed or blocked synchronization jobs are no longer counted as though they were active jobs stuck in progress.

We also reduced noisy browser-protocol warnings and updated the agent’s model-library integration. ( #1428, #1466, #1596)

Website delivery and infrastructure

The website rebuild includes changes to DNS, CDN delivery, reverse-proxy configuration, caching, media publication, analytics, and search metadata.

These support the redesigned site and its media. Performance targets in the website plan are not presented as measured results. ( #1437, #1442, #1443, #1447, #1485, #1489, #1499)

Plans stay separate from released features

The website and billing plans were updated alongside their implementation.

Unimplemented sections, demonstration scenarios, and proposed follow-ups are not included simply because they appear in those documents.

These notes describe the implemented product changes and the paid-plan launch confirmed for this release.

Have a feature your team needs? Tell us at founders@heyhyper.ai.