New features, improvements, and fixes shipped to PixTwins.
PixTwins now installs to your home screen on iPhone and Android, runs full-screen like a native app, and can push notifications to you when an agent finishes a long task or needs your approval. Every primary screen has been tuned for thumb-only use.
A new Templates page surfaces the full catalog of prebuilt twins and bundled teams. Pick a template, hit install, and you land in a fresh chat with your agent already onboarded — no instructions to write, no setup to do.
PixTwins now requires an active paid plan. The free tier has been retired. All paid plans continue to include BYOK (bring your own provider key), so power users with their own Claude / Gemini / OpenAI key can still keep usage costs low while paying just the platform fee.
Usage now tracks against your actual billing period instead of the calendar month, so mid-month signups get exactly one full allowance per cycle. Full refunds correctly end access on the refund date.
Patch release: when the CLI is run while signed out, it now exits with code 1 and reports an authentication error, instead of looking like a server bug. Update with `npm install -g pixtwins@latest`.
The pixtwins command-line tool now mirrors what you can do in the web app. Browse and edit agents, memories, integrations, automations, skills, and outputs from your terminal. Run `npm install -g pixtwins@latest` to upgrade.
Fixes a regression where signing in to the `pixtwins` command-line tool would silently fail on the latest Chrome and Edge. Run `npm install -g pixtwins@latest` to pick up the fix; existing sessions keep working.
Every PixTwins plan now leads with access to the most powerful models from Claude, Gemini, and OpenAI -- one subscription, every frontier model, no separate per-provider accounts. The pricing page, landing page, and in-app model picker have all been updated to make this front and centre.
Plans rebuilt around a monthly token allowance you can spend across Claude, Gemini, or OpenAI. Heavier models burn more tokens, lighter ones burn fewer — same pool, your call. Bring your own API key on every plan. Pro now includes priority support; the new Max plan includes expert setup and a custom-branded portal.
Switch the engine that powers your workspace between Claude, Gemini, and OpenAI. The chat header always shows which model is running, and the provider selector now lives next to the chat box.
Image creation now starts with the faster image generation path for everyday creative work, while keeping a high-quality fallback when needed.
Switch the engine that powers your workspace between Claude, Gemini, and OpenAI. The chat header always shows which model is running so you can pick the right one for the task.
Agents now choose compatible actions more consistently when you create a team, install a bundle, connect apps, or ask for code, data, media, and monitoring work.
When you connect an app like Klaviyo, Gmail, or Slack, your agents now receive a better-balanced set of available actions for the current task. Bundle-installed agents can also pick up relevant new app connections later instead of being locked to the apps the bundle was originally built around.
Sending a message no longer makes the chat area briefly disappear and reappear. The message list now stays stable through the whole send — from the moment you hit enter to the agent's first word — instead of blanking out and redrawing a few times along the way.
Made a typo, forgot a detail, or want to rephrase a request? Hover over any message you sent and hit "edit". Update the text and your agent replies to the new version — everything that came after the edited message is cleared so you get a clean rerun instead of stacking on top of the old thread.
Drive your PixTwins workspace without leaving the shell. Install the CLI, sign in through your browser, and you can run one-shot prompts, drop into an interactive chat, or pipe prompts into scripts and cron jobs — all using the same agents, memories, and integrations as the web app.
Once your fal.ai key is in Settings, your agents get three new generation tools. Ask for an image and they can call Flux; ask for a video and they reach for Kling, LTX, Wan, or Veo 3; ask for a soundtrack and they call Stable Audio or MiniMax Music. Outputs land in your Outputs tab automatically.
Add your fal.ai API key in Settings to plug into hundreds of open models for image, video, audio, and 3D generation. Sits alongside your existing Anthropic, Gemini, Higgsfield, and ElevenLabs keys.
Our image tool now runs on a newer model with crisper text rendering and more accurate scene composition. Posters with headlines, product mockups, and any prompt that asks for text-in-image should look much cleaner now.
Each workspace can now choose which provider runs the main agent loop. Use the toggle in the top bar to switch between Claude (premium quality, our default) and Gemini (efficient mode with very long context). Your choice persists per workspace.
Opening the PixTwins app from inside your Shopify store admin now correctly connects the store to your workspace. Previously the OAuth round-trip would finish but silently fail to save the connection, leaving you on the chat page with no store linked.
Fixed a UI glitch on the Integrations page where the Shopify connect option could appear twice when other integrations were already connected.
Hardened the subscription system to prevent duplicate charges and keep subscription status accurate during delivery hiccups.
Every workspace gets a real working environment where agents install libraries, log into sites, and run watchers that keep going between sessions.
Subscription plans are now active. Start with a 7-day free trial on any paid plan — no charges until the trial ends, cancel anytime. Manage your subscription from Settings.
Creating an always-on automation is now a guided 4-step flow. Pick the trigger type, configure it, describe the action, and launch — your agents handle the rest.
Added a common-questions section to the landing page so visitors can quickly get answers about what PixTwins does, pricing, data safety, and team workspaces without having to email support.
Mentioning a specific agent in chat is now faster and clearer. You can see which agent will receive your message before you send it, quickly reply to any agent with one click, and mention detection is more accurate.
Invite collaborators to your workspace. Share access with teammates so they can view agents, chats, and memories without needing to set everything up themselves.
The dashboard now shows 14-day token and task trend charts so you can see how your agents are being used over time. You can also schedule daily or weekly digest emails, and search through your memories instantly.
Memories are now stored as a connected knowledge graph — relationships between facts are detected automatically, similar memories are grouped into communities, and context selection prioritizes the most relevant clusters. Plus a new Activity page and an improved onboarding wizard.
You can now ask any agent to create a reusable custom skill mid-conversation. The skill is saved to your workspace and automatically added to that agent. The agent builder also has a cleaner editing experience.
Always-on chats now have a visual workflow editor that shows exactly what your automation does — trigger, actions, conditions, and delays — as a node graph you can view and adjust.
Voice conversations now route through the full agent pipeline. Your agents can use tools, create automations, delegate to specialists, and save memories — all via voice.
Talk to your agents in real time with the rebuilt voice conversation mode. Full-duplex audio — speak naturally and hear your agent respond instantly.
Agents can now get individual product details, create new products, view and update orders with fulfillment tracking, and look up customer profiles.
Customize which notifications you receive with per-type preferences. New notification types for automation results and system announcements.
The desktop app now checks for updates automatically and installs them in the background. No more manual downloads.
Rate and review community skills in the marketplace. A new featured section highlights top-rated skills.
Your thumbs up/down feedback on agent messages now persists across page reloads. Feedback is stored per message and rate-limited.
Toggle voice mode in any chat to talk with your agent in real-time. Your agent listens, thinks, and speaks back instantly — a natural back-and-forth conversation.
Notifications now cover more events across the platform, with granular per-type preferences so you control exactly what shows up. The desktop widget also includes notification counts in its badge.
Agents can now get individual product details, create new products, view and update orders with fulfillment tracking, and look up customer profiles. Expanded store management scopes for complete control.
Customize which notifications you receive with per-type preferences. New notification types for automation results and system announcements, plus automatic cleanup of old notifications.
The desktop app now checks for updates automatically and installs them in the background. No more manual downloads -- just click "Update now" and the app restarts with the latest version.
Rate and review community skills in the marketplace. A new featured section highlights top-rated skills, and sort controls let you browse by popularity, rating, or recency.
Your thumbs up/down feedback on agent messages now persists across page reloads. Feedback is stored per message and rate-limited to prevent abuse.
A major visual upgrade to the entire app. Chat messages now render rich formatting -- headers, bold, code blocks with syntax highlighting, tables, and more. Loading states are smoother, components are more polished, and keyboard accessibility is improved across the board.
The desktop app now stays running in your system tray when you close the window. Your widget and always-on agents keep working in the background. Plus, future updates install automatically.
Desktop app users get a sticky floating widget that stays on top of all windows. Chat or speak to your agents instantly — with full screen awareness, voice responses, and one-tap quick actions.
Slack, Gmail, Google Calendar, and Google Analytics now connect directly to your workspace — faster, more reliable, and no middleman. Agents get 24 new native tools.
Connect multiple Shopify stores to a single workspace. Manage all your stores from one place, set a default store for agents, and switch between stores when needed.
PixTwins now automatically detects what plugins, integrations, and tools your workflows need -- and helps you install them in one click. A unified capability registry brings together MCP servers, integrations, skills, and bundles into a single browsable catalog.
Agents now remember the full conversation history across long chats, and have awareness of other active work in your workspace. Stream reconnection is also more reliable.
See your token usage at a glance with the new usage badge, create and share skills from a community library, and benefit from stronger billing enforcement and expanded test coverage.
Generated images, HTML previews, and other artifacts now reliably persist across page refreshes. Agents can also run longer on complex multi-step tasks without stopping prematurely.
Comprehensive security audit and fixes across the platform. Stronger input validation, improved webhook protection, better error handling for long-running tasks, and startup health checks ensure your data and workflows are safer than ever.
Agents can now directly manage your Shopify store — browse products, check orders, create landing pages, edit theme templates, and analyze sales data. Connect your store and let your digital twin handle the day-to-day.
Major intelligence upgrade — agents can now analyze images and screenshots, generate audio briefings, and respond to your voice. Plus significantly faster routing, upgraded web search, and reduced costs across the board.
Agents now automatically recover from transient failures instead of crashing. A new lifecycle system tracks agent state from start to finish, and workspace policies let you control agent behavior with declarative rules.
Agents now automatically learn from your connected apps. Gmail, Slack, and GitHub data is indexed in the background so your agents know about your work without being told. Plus, usage tracking, template upgrades, and always-on reliability improvements.
Agents can now conduct multi-source research with parallel search and synthesis. They also reason more carefully through complex multi-step tasks, and workspace memories are automatically deduplicated in real-time.
New users now get a smoother start with generalist agents, template previews, and smarter guidance. Plus, agent performance is now tracked and visible so you can see how your team is doing.
Agents now preserve structured context during long tasks, so they remember goals, progress, and key decisions when continuing work. Integration tools also load faster with parallel fetching.
Every time an agent learns and updates its own instructions, PixTwins now saves a versioned snapshot. You can view the full version history from the agent editor and roll back to any previous version with one click.
When you correct an agent, the system now automatically saves what NOT to do as an anti-pattern memory. Agents learn from mistakes and actively avoid repeating them.
You can now configure each agent's communication style with tone, verbosity, and formality controls. Customize how your agents talk -- from concise and technical to friendly and detailed.
You can now see what each tool returned during agent execution instead of just "Using tool..." with no feedback until the final response arrives.
Agents now automatically consolidate their instructions when they grow too long from self-learning, keeping them focused and contradiction-free. Learnings are also shared across your team as workspace memories.
A major upgrade to agent intelligence and user control. Agents now think deeper on complex tasks, learn from your feedback, and can be individually tuned with model preferences and cost limits.
A quality-of-life update addressing the most-requested user feedback: multiline chat input, agent team grouping, direct agent chat, workspace renaming, improved empty states, better output visibility, and enhanced API key management.
New pre-built workflow for traders: a 5-agent team that covers the full trading lifecycle from market research to algo deployment. Includes real-time data tools, strategy backtesting, risk management, and automated trade execution.
Major platform expansion: agent performance dashboard with leaderboard, skill marketplace for community sharing, upgraded outputs gallery with search and bulk actions, plus real-time financial data tools for traders and analysts.
Your Anthropic API key now powers your entire workflow. Add your key in Settings and every agent call, routing decision, and tool execution uses your key instead of PixTwins credits. Work without token limits.
Skills are now fully functional. When you assign a skill to an agent, its specialized instructions are injected into the agent's behavior. You can also create your own custom skills from the Skills tab, via chat, or inline in the agent editor.
Major upgrade to the orchestration engine. Agents now route tasks to the right model based on complexity, track their own performance, automatically summarize long conversations, and consolidate related memories into higher-level insights.
Agents can now search across all your past conversations in a workspace, finding relevant context from previous chats and memories. The main coordinator agent has been rebranded to PixTwin with an updated execution style.
The Email Marketing template now includes three new specialized agents for DTC brands — a retention strategist that audits flows and plans A/B tests, a monthly performance reporter that pulls Klaviyo + Recharge data into Google Sheets, and a QA auditor that catches grammar errors and broken links weekly.
Major upgrade to the agent execution engine. Agents now run on a more capable foundation with smarter delegation, adaptive thinking effort, budget safety, and enhanced self-learning from your feedback.
Build structured "When X happens, Do Y" automations that execute without AI cost. Describe what you want in chat, and the agent guides you through trigger selection, field mapping, and confirmation with interactive UI elements -- all inline in the conversation.
Upgraded our agent engine for more powerful autonomous workflows. Agents now have true agentic loops with tool use, error recovery, and multi-step planning built in.
Always-on agents can now define pre-filters on their triggers so they only process events that match specific criteria. This reduces noise and saves your plan budget by skipping irrelevant events before they reach the agent.
Fixed a bug where OAuth callbacks would redirect to the wrong URL after connecting an integration. OAuth redirects now work correctly in all deployment environments.
All integration credentials are now protected with additional security layers. Your connected app credentials are safer than ever with no changes needed on your end.
Code execution results now stream directly into the chat interface. Results are displayed inline so you can see exactly what your scripts produce without leaving the conversation.
Agents can now modify their own automation configurations through a new tool call. This means you can ask an agent to change its schedule, update trigger filters, or adjust its behavior without visiting the settings page.
Workspaces now support multiple independent chat threads. Create task chats for one-off work or always-on chats for continuous monitoring. Each chat maintains its own conversation history while sharing workspace-level agents and memories.
When a conversation grows very long, the system automatically summarizes earlier context to keep things running smoothly. This keeps conversations functional indefinitely without losing important details from earlier in the thread.
Introducing 24/7 monitoring agents that stay active and process real-time events from your connected apps. Set up a Slack monitor, Gmail watcher, or analytics alerter that works around the clock without manual prompting.
Long-running agent tasks now automatically continue where they left off when they reach a natural breakpoint. A continuation summary is generated and the next request fires seamlessly so complex tasks run to completion.
Agents can now write and execute Python and Node.js code in a secure environment. Each execution is fully isolated with automatic cleanup for safety.
A multi-agent workflow for end-to-end video creation. From research and brief writing through storyboarding, asset generation, and final editing -- your agent team collaborates to produce complete video content.
Each agent now has its own set of allowed toolkits. This means your CRO agent only sees analytics tools while your email agent only sees Gmail -- reducing confusion and improving accuracy.
Use your own Claude, Gemini, or OpenAI API keys for unlimited usage beyond your plan's token allocation. Keys are stored encrypted and used transparently -- you just pay your plan fee plus your own API costs with zero markup.
A persistent memory system that lets your agents remember brand guidelines, preferences, and past decisions. Memories are indexed for fast intelligent semantic search during agent execution.