What's new in Flashbag.

Fixed

• Profile updates now only change what the update changes, so they no longer fail or report conflicts because of your own files. For each conflict, choose whether to keep your version or use the new one.

New: Agent Services

• Agents can now keep programs running after a reply finishes. Ask an Agent to build a website or web app, then open its preview in your browser. Previews update automatically as the Agent changes them.
• Manage these on the new Services page: start, stop, open, check their status and recent output, or remove them without deleting project files. Organize them into Service Groups, and ask Gordon for help with them in any Profile.
• Service cards in chat show current status and controls, even in older conversations. Services also appear beneath their Agents in the Agent list.
• Open web Services on the computer running Flashbag, or privately from your other devices through Tailscale. If you use Flashbag through Launcher, previews need Tailscale. Services stop when Flashbag closes and can be started again later.

New: Permissions and questions

• Choose how much an Agent may do on your computer before it asks you: for every Agent in Settings, for one Agent on its page, or right from the message box. Claude Code and Codex offer their own modes, from asking before every edit to full access. Flashbag recommends Auto, which only asks about actions that look risky.
• When an Agent needs permission, it asks in the chat and shows what it wants to do and why. Allow it once, always allow it, or deny it and say what to do instead. Always allow applies to every Agent using the same AI tool.
• Tasks run with no one there to answer, so anything that needs approval is denied. Open View Activity on a reply to see every decision made.
• Agents can also ask you questions in the chat. Pick a suggested answer or write your own. If an Agent asks for something private, like a password, your answer reaches it but is never saved.

New

• Claude Haiku 5.5 is available, and Flashbag now uses it by default for its background memory work with Claude.
• Send now sends a queued message right away, stopping the current reply but keeping what it wrote so far.

Improved: A unified look

• Flashbag has a new, unified design. Buttons, fields, menus, dialogs and notices look and work the same on every page, in light and dark themes, with better contrast throughout.
• The Agents, Skills, Services and Tasks pages share one layout: search, filter by Agent or Group, and select several items to act on at once. The selected row stands out.
• Settings are regrouped into clearer sections with short explanations, and Activation and the Subscription page are redesigned.
• An Agent's page opens with its role, memory and live activity, shows the AI tool it runs on, and marks settings changed from your defaults. Agents and Profiles can use any custom color.
• In chat, usage for the current session and today's estimated cost sit beside the message box, notices above it stack in one place, and smaller headers leave more room for conversations, especially on phones.
• Forms show every problem at once, next to its field, and have a clear place to add files.

Improved: Memory

• Agents now choose what's worth remembering while they work and save it as Signals. This replaces the helper that picked out facts after every reply. Facts from earlier versions stay available to Agents, marked as possibly out of date, but no longer load when a conversation starts.
• The Memory at the start setting now sets a firm limit on how much an Agent recalls when a conversation starts, so Agents with long histories start faster and cost less. Flashbag also warns when a level is a poor fit, such as None for chat.
• When one of an Agent's identity files grows too large, the chat tells you, and Optimize asks the Agent to tidy it up. View and edit these files on the Agent's new Identity tab.
• Gordon now writes down what he learns about you in a note every Agent in the Profile reads, so they all know who you are. You can edit it on any Agent's Identity tab.

Improved: Tasks and Skills

• Organize Tasks into Task Groups. The redesigned Tasks page shows each Task's Runs beside it, and you can filter results across all past Runs.
• Add a Task or Skill by describing it to Gordon, who prepares it for your approval, or fill it in yourself. Gordon can also import a Skill you already have from a folder or zip.
• Editing a Skill shows its instructions and its other files. Drag Skills to reorder them or move them between Groups.
• Agents can create Skills for themselves, and update or repair the Skills they use when you ask.

Improved

• Claude Code, Codex and Pi are now called AI tools. A Dashboard card shows each one's version, updates and sign-in, and Settings lets you turn them on or off and choose which installed copy each uses. Signing in works the same way for each.
• First launch is clearer: turn on the AI tools you want and see why Continue isn't available yet. If Gordon needs you to sign in, you can do it right where you start chatting.
• The Dashboard's Usage card compares this week, month and year with the previous period, and What used it breaks usage down by model, Agent or role.
• Flashbag is faster: chats with long histories open quicker, and Flashbag loads faster in a browser, especially over Remote access.
• Reorder Agents by dragging them on any page that lists them. The Agent list shows each Agent's role, or today's tokens when it has none, and Agent Groups remember whether you left them expanded or collapsed.
• Flashbag's own updates sit at the top of the Dashboard.
• Gordon can manage Agents, Tasks, Skills and Services, and their Groups, in any Profile without you switching Profiles first.

Fixed

• Sent messages no longer appear to fail during long replies, replies no longer stall on their last few words, and a queued message no longer gets stuck when a reply fails unexpectedly.
• Messages whose reply failed or was stopped now stay in the chat after reloading, and the Agent receives them with your next message.
• Gordon no longer sends a greeting to a chat left open overnight with no one there.
• Agent activity indicators clear reliably when work finishes.
• Agents with Memory at the start set to None no longer read their Memory on their own. They read it only when you ask or agree.
• A rare problem with an Agent's saved memory no longer makes every message to it fail. Flashbag now repairs it automatically.
• Quick edits to an Agent's identity files are less likely to overwrite each other.
• Claude Code and Codex no longer load an Agent's instructions twice, so conversations start with fewer tokens.
• A Task that's running when you quit Flashbag continues when you reopen it, instead of being canceled.
• Background sign-in checks no longer interrupt working Pi chats or report misleading setup errors. Connecting a Pi provider no longer looks like it failed when Pi can't load its model list.
• On macOS, Codex chats no longer fail with "Too many open files".
• Adding files no longer fails on large disks with plenty of free space. Uploads now only need 5 GB left free, and the message says how much is needed and how much is free.
• Long file names no longer push buttons out of view when adding files, and dialogs can't be closed by accident while saving.
• In Flashbag Desktop, links in chat and the Remote access and Tailscale addresses now open in your browser.
• When Flashbag is read-only because your access has ended, the chat says so and links to Subscription, instead of quietly refusing your messages.
• Failed Tailscale sign-ins explain the problem, and Retry starts fresh.

New

• GPT-6.1 Sol is available in the model picker and is now the default for Agent replies and Gordon when Codex is enabled.
• Claude Sonnet 5.5 is available in the model picker and is now the default for Claude helpers.
• Choose an App name and Icon color in Settings → General to recognize Flashbag in your browser tabs, including through Remote access. Tabs show your chosen name alongside the current Profile, with the Flashbag icon in your chosen color.

Improved

• New installations guide you to sign in to Claude or Codex before meeting Gordon, with step-by-step instructions on the Dashboard and in Gordon's chat. When Flashbag runs through Flashbag Launcher, the instructions point to the Launcher's Open Terminal, the only terminal that can sign Flashbag in.
• Until onboarding finishes, the Dashboard shows only your next step instead of empty statistics.
• Pi Agents now use the MCP support built into Pi 1.0 and need Pi 1.0 or newer. An Agent's first chat no longer has to download an extension, and Flashbag removes the one it installed for Pi before.
• Codex Fast mode now clearly shows its 2.5× subscription usage, separately from estimated costs.
• Removed GPT-5.4, GPT-5.4 Mini, and GPT-5.5 from the Codex model picker. Settings and Agents using these models now follow current defaults.

Fixed

• Flashbag always opens, even when Claude, Codex, or Pi is signed out or unavailable. Chats and Tasks that can't run tell you why, so you can sign in or choose another model.
• Tasks that start just after you open Flashbag now wait a moment while it checks your sign-ins, instead of failing. Pi Agents no longer need a restart to work.
• When a scheduled Task can't run, its history now shows why instead of it silently missing its turn.
• Fixed Pi fact capture missing facts the Agent had already saved during the chat, which could cause near-duplicates.
• When a Task fails because you're signed out, it now offers to sign you in.
• If one Profile can't load, the rest of Flashbag still opens, with an option to try again.
• Fixed Flashbag sometimes hanging at startup, or not opening after a Skills folder was changed outside Flashbag.

Important: First startup after updating

• First startup may take longer while Flashbag prepares earlier chat activity for faster loading. Leave it running until startup finishes. Your conversations are preserved.

New

• Let Claude use the connections linked to your Claude.ai account. Enable Use Claude.ai connectors in Settings → CLIs. It is off by default; enabling it may delay the start of replies.

Improved

• Follow work delegated to Claude and Codex Subagents in View Activity, with their tool steps grouped under the call that started them. Chat, Task, and usage totals now include their tokens and costs, with a separate Subagents breakdown.
• Codex replies stay in progress while their Subagents finish, and Stop stops that work too. Pauses during Codex replies now appear in View Activity with their duration.
• Paste images from your clipboard directly into chats.
• Long chats and Task results open faster. Task history now has pages and remembers your filters and reading position when you return from a result.
• Saved replies and copied answers now contain just the final response. Progress messages remain available in View Activity.
• Agents look up saved facts and conversations directly, simplifying memory settings by removing separate search-model choices. Existing memory is preserved.
• After an update, Agents automatically pick up the latest instructions and tools while keeping their saved conversations and memory.
• Flashbag uses less disk space for troubleshooting details. Debug mode is available again in Settings; oversized diagnostic files show a clear message instead of freezing the app.

Fixed

• When Codex needs more information, it asks in its reply instead of opening a question dialog Flashbag cannot show.
• Images generated with Codex are saved and shown in chats and Task results, even when the Agent forgets to include an image link.
• Conversations, including Gordon's, stay visible when Claude, Codex, or Pi is unavailable. Clearer guidance explains how to get replies working again.
• Claude no longer attempts to open unsupported interactive dialogs. Its separate scheduling tools are disabled so scheduled work stays in Flashbag Tasks.

New

• GPT-6 Sol is available in the model picker and is now the default for Agent replies and Gordon when Codex is enabled.
• GPT-6 Luna is available in the model picker and is now the default for helpers when Codex is enabled.

Improved

• Flashbag no longer stops agents after 15 minutes by default. In Settings → Agents, choose Unlimited or a preset time limit. Updating resets any previously saved limit to Unlimited once, so choose a new one if needed.
• Usage estimates now reflect OpenAI's updated GPT-5.6 Sol rates and Fast mode prices for GPT-5.6 models and GPT-6 Astra.

New

• Claude Opus 5.5 is available in the model picker and is now the default Claude model for Agent replies and Gordon.

Fixed

• Chats now use Claude's default models when Claude is enabled and Codex is disabled. Custom model selections are preserved.
• Automatic greetings now show how long they took, including first-time setup, new Profiles and Agents, and Gordon's Agents and Skills side panels.

Important: Remote access needs setting up again

• Before updating remotely, make sure you have another way to access the computer running Flashbag. Remote access needs reconnecting after this update, and your old Flashbag address will no longer work.
• After updating, open the desktop app on that computer, choose Settings → Remote Access → Connect to Tailscale, follow the sign-in, and save your new address.

New: Run Flashbag with Launcher

• Meet Flashbag Launcher, a separate app you download and install to give your Agents their own space on your computer. They can install tools and work on files inside that space, helping keep their changes away from your everyday files and apps. It's a good fit when you want to run Flashbag on a computer you also use for other things.
• Flashbag Desktop is designed for a dedicated Mac or Linux computer. It runs directly on the device, giving your Agents access to work across that computer. Choose Desktop when you want that freedom, or Launcher when you prefer to keep their work contained.
• You can switch from Desktop to Launcher on the same computer and keep your Agents, memory, conversations, and settings. Download and install Flashbag Launcher separately, then update Desktop, let any running work finish, and quit it before starting Flashbag through Launcher. You'll need to sign in to your AI services again the first time.
• With Launcher, Flashbag runs in the background and you use it from your browser. Launcher handles installation, starting, stopping, and updates, including guided setup on Windows. Your Flashbag data stays on the hosting computer.
• Agents running through Launcher come with tools for browsing websites and working with documents, images, and media, so more Tasks work without extra setup.
• Flashbag lets you know when an update is available and shows what's changed. When running through Launcher, you can download the update while you work and apply it when you are ready to restart.

New: Built-in Tailscale connection

• Connect Flashbag to Tailscale directly from Settings to use it from another computer or phone. There's no longer a separate Tailscale installation to manage on the computer running Flashbag. Your other devices still need Tailscale connected to the same network.
• Connection status, addresses, and password controls are together in one place. Choose an address name that's easy to remember, with clear guidance when sign-in or device approval needs your attention.
• Your existing Remote access password is kept.

Improved

• Codex sign-in now gives you a code to enter in your browser, making it easier to connect from the desktop app or another device through Remote access.
• Claude sign-in provides a command to copy and clear instructions for where to run it. Flashbag checks for completion automatically, and Settings now lets you reconnect Claude or Codex when needed.
• Updates now make extra checks that the download is the right Flashbag release for your computer before installing it.

Fixed

• Chats can recover when Claude, Codex, or Pi can no longer reopen an earlier session. Flashbag starts fresh with the Agent's usual memory preload and retries your message, while keeping your saved conversation.
• Browser access now offers Reload Flashbag when an update leaves the page out of date or it cannot connect.
• Failed uploads show clearer errors that you can dismiss before trying again.
• Correcting your computer's clock and checking your Subscription again can now restore access without resetting your activation.
• Restarting during first-time setup keeps Gordon's guided setup available instead of creating an Agent before you are ready.
• Profile update reviews no longer briefly show an earlier comparison while a new one loads.
• If another copy of Flashbag is already using your data, startup now explains what to close so you can continue.

Compatibility: Claude sign-in

• Claude now keeps its Flashbag sign-in and settings separate from other Claude use on your computer. Existing Claude users need to sign in again using the command shown in Flashbag. Your saved Flashbag conversations and memory are kept. Connections added through claude.ai no longer carry over into Flashbag.

Important: First startup after updating

• First startup may take a while as Flashbag cleans up oversized diagnostic event logs. Leave it running until startup finishes. Your conversations and images are preserved.

New: Redesigned chat

• Chat now puts the final reply at the centre. Minimal activity is the default, collapsing thinking, tool activity, and intermediate replies when work finishes. Open View Activity to inspect the steps, or choose Full activity in Settings to keep them visible.
• Reply controls sit together beneath each response: copy the final answer, check elapsed time and token usage, inspect activity, or ask Gordon for help. Choose your own message colour and avatar in Settings.
• Browse images in galleries with thumbnails, keyboard and swipe navigation, and downloads for individual images or the whole gallery. File attachments show clearer names, locations, and file types.

New

• Edit Skills individually or in bulk, organize them into Groups, and filter by Skill Group, Agent, or Agent Group. Skills assigned to Agent Groups follow Agents as they join or leave.
• Set app-wide Preload, model, and reasoning defaults. Agents using defaults follow later changes automatically, with Preload and Pipeline customizable independently for each Agent.
• Experimental Pi support connects Agents and pipeline helpers to ChatGPT or OpenRouter. It is off by default and intended for advanced testing; enable it in Settings → Experimental, then configure it under CLIs.

Improved

• Change model and reasoning choices directly in chat, and Preload on the Agents page. Choices save immediately, and Reset restores your configured defaults.
• Checking for a Profile update automatically prepares a review with grouped Agents and Skills, browsable folders, full file comparisons, and conflicts highlighted before approval.
• Turn Claude Code and Codex on or off, or choose between installed copies, without restarting Flashbag. Pi offers the same controls when enabled in Experimental settings.
• Zoom controls replace layout density, letting you adjust the size of the app on each computer.
• Internal changes in preparation for upcoming Windows and server support.

Fixed

• Fixed image-heavy chats hanging and causing excessive event-log growth.
• Stopping a Codex reply or encountering an error in one conversation no longer interrupts other Codex work. Refreshing sign-in status also leaves running work alone.
• Switching between long chats uses less memory while preserving your reading position and access to older messages. Image previews no longer shift reopened chats as they load.
• Flashbag stays responsive during startup while Remote access connects.
• Application updates now reject unsigned releases before downloading them.
• Skills created or edited outside Flashbag now refresh reliably. Damaged settings no longer hide healthy Skills or accidentally grant access, and interrupted moves preserve access restrictions.
• Sign-in failures during chat now appear across Flashbag, and Refresh clears chat warnings when access is restored.
• Updates to Claude Code, Codex, and Pi now take effect without restarting Flashbag. Refresh can recover interrupted updates, and canceled updates no longer continue in the background.

Compatibility

• Update Flashbag to receive Profile updates that include Agent Group Skill assignments or saved Skill Group names and order.
• During setup, choose at least one command-line tool to use with Flashbag. Claude Code and Codex are available by default; Pi requires enabling experimental support. Sign-in problems no longer prevent you from opening Flashbag.

New

• GPT-6 Astra is available in the model picker, above GPT-5.6 Sol, with Low through Ultra reasoning and Fast mode.
• The Dashboard now provides another way to check for updates, including through Remote access.

Fixed

• Resumed Codex chats keep live progress and tool activity on the correct message, and Gordon's tools use the Profile captured for each request.
• Reloaded Remote access chats show when work is still running, recover finished replies automatically, and report when no reply was saved.
• Codex image generation no longer causes chat turns to fail.
• Gordon now accepts uploaded files from every selected Profile while continuing to keep his files in the Default Profile.
• Flashbag's Codex sessions no longer appear in ChatGPT or ordinary Codex CLI history. Existing Codex users may need to sign in once through Flashbag.
• Voice dictation now works in Linux AppImage installs.
• Selecting Agents works again after removing the Selected Profile.
• Ask Gordon is now available on failed turns to help diagnose problems.

New

• Run an Agent's saved Tasks directly from chat with /task: <name>, including clearly labeled slash-menu suggestions alongside Skills, live progress, links to completed runs, and each Task's saved result-posting preference.

Improved

• Task names must be unique within a Profile; existing duplicates are flagged for repair and cannot run manually or on schedule.

Fixed

• Send Fresh in the chat composer now starts a new conversation before sending the draft.

New

• Voice dictation is now available in the desktop app and over HTTPS Remote access, with downloadable local English or multilingual speech models, live text previews, and offline use after installation.
• Claude Fable 5.1 is available in the model picker.
• New Agents prefer GPT-5.6-Sol whenever Codex is available, while Claude-only setups continue using Claude Opus 5 as default.
• Dashboard usage can now switch between the selected Profile and all Profiles, remembers that choice, and attributes Gordon's usage to the Default Profile instead of repeating it across Profiles.

Improved

• Chat now lets you on the fly switch models and reasoning levels without first going to the agents page.
• Codex Fast mode delivers responses 1.5× faster while using 2.5× as many ChatGPT credits.
• Codex support has been substantially upgraded, with reliable conversation continuity, live replies and tool activity, and accurate usage reporting.
• When a Claude Code or Codex session expires during chat, Flashbag now offers direct sign-in and a connection check instead of raw CLI errors.
• Remote access now uses Tailscale HTTPS automatically when a certificate is available, while old HTTP addresses redirect on the same port and HTTP remains available as a fallback.
• Profile creation now sits directly beneath the Profile list, while the Dashboard header shows the current Profile and its management actions.

Fixed

• Tasks started by Gordon now run to completion instead of being canceled as soon as they begin.
• Claude Code and Codex setup now keeps sign-in status visible alongside available updates, explains when an installation source does not offer the newest release yet, and replaces technical update output with clear outcomes.
• Agent avatars selected from local image files now render in browser sessions.
• Gordon's activity details and Debug Copy now remain complete after working in another Profile.
• Activation reset screens no longer show an empty secondary button after status updates.

Compatibility

• Flashbag for macOS now requires macOS 15.5 or newer.

Fixed

• Remote access now opens reliably instead of remaining on “Preparing Flashbag…”.

Improved

• Activation offers clearer options for starting a Trial, freeing another computer, or requesting an email reset.
• Claude Code and Codex setup is now guided at startup, remembers your choices, and never installs anything automatically.
• Agent reasoning settings now show Flashbag's recommendation for each model.
• Narrow windows leave more room for Agents and Skills.

Fixed

• Fixed computers occasionally being misidentified and entering read-only mode. Existing Flashbag homes must activate once after updating; local data is preserved.
• Gordon's onboarding now remains available across restarts while access is being activated.
• Remote access now detects outdated browser tabs after Flashbag updates and asks you to reload.
• Flashbag now catches Claude Code and Codex installation, update, and sign-in problems before startup.

New

• Linux is now officially supported.

Improved

• Linux now explains how to finish setup when a required system component is missing instead of silently failing to open.

Fixed

• Profile updates performance improvements.

Improved

• The selected Profile is now clearly marked in the sidebar, with hover feedback for other Profiles.

Fixed

• Remote access now stays disconnected when Flashbag is read-only.
• Profile updates now show progress and errors, cancel safely while preparing, and recover Profile navigation without requiring an app restart.
• Flashbag now uses the same Claude Code and Codex installations as the user's Terminal, handles official, Homebrew, and npm releases separately, and identifies shadowed copies.

Fixed

• Fixed several reliability issues.

New

• Switch between the first nine Profiles with ⌘1–⌘9.

Improved

• Profile icons now reveal their names, and Profile and Agent name fields have more room.

Fixed

• Improved the reliability of Application updates on macOS.

New

• App updates now appear in a bottom bar, including before activation.

Improved

• Agents link files more reliably for Browser users.

Fixed

• Fixed Gordon's identity, Profile context, cross-Profile access, and management tools.

Removed

• OpenClaw and Telegram importing are no longer available (better importing is coming in a future update).

Improved

• Improved Subscription and Trial reliability.

New: Multiple Profiles

• Create and switch between independent Profiles, each with its own Agents, Groups, Skills, Tasks, history, and output.
• Give Profiles and Agents their own images (or emoji as before). Large photos are reduced automatically, including uploads through Remote access.
• Copy Skills between Profiles and choose exactly which Agents can use each Skill.
• Blank Profiles open with Gordon, who can help set them up.
• Initial release of shared profiles for select Customers. Pull curated Profiles into independent local copies, review available updates, and apply them without replacing local-only work. Gordon guides you through any conflicts.

New: Subscriptions and Trials

• Start a card-free Trial by email, or activate a paid Subscription from the new Subscription page.
• Existing data stays readable when access ends or cannot be confirmed. Flashbag pauses new work and Remote access, then resumes automatically when access returns.

Improved

• Recover broken Agents and Skills from their pages with Finder access and safer confirmation.
• Linux now uses the Flashbag icon in launchers and windows.

Improved

• Significant internal re-organization in preparation for future features.

Improved

• Chat on iPhone no longer zooms in when you tap the message box.
• The top bar and the message box stay in place instead of sliding behind Safari's toolbars.
• Long agent and task lists now scroll on phones.

New: Small screens

• Flashbag now works on phones and narrow windows, making remote access from a phone usable.
• The window can now be resized down to 320 by 480.

Improved

• Task runs now show a cost for models whose CLI reports none, including every GPT model.
• Gordon can now retry a skill change that partially failed.

New: Agents and Groups

• Organize agents into collapsible, reorderable Groups and subgroups across Agents, Chat, and task agent pickers.
• Create and edit agents from the Agents page, including their name, personal or business context, role, emoji, color, and Group; newly created agents open Chat and begin context-aware guided setup.
• Agent deletion now uses two confirmations, shows exactly which identity, memory, transcript, session, and organization data will be removed, preserves saved tasks and agent outputs, and refuses deletion during an active task run.

Improved

• All available referenced files now get download cards even when they cannot be previewed, and small text file previews continue loading beyond the fourth referenced file.
• Browser uploads now explain when they are unavailable because the destination volume is at least 90% full.
• Usage cost tracking now recognizes Claude Sonnet 5 and GPT-5.6 models.
• Gordon now reports when a skill change succeeds but its CLI links fail to refresh, including guidance to reconcile them.

New: Add files from anywhere

• Use the new + control in Chat, Gordon, and Tasks to select multiple files or folders already on the Mac.
• Remote access browser sessions can browse files on the Mac or upload up to 20 files from the current device, with visible progress and the option to resume by reselecting a file after a refresh or interrupted connection.
• Drag and drop files on chat now works with remote access browser sessions.

New: Visual file previews

• Files referenced in messages now appear below them: images and text can be previewed, PDFs get file cards, and originals can be opened on the Mac or downloaded from a remote access browser session.
• Messages load at most 20 distinct file references and disclose any remainder.
• Works for both your messages and the agents.

Improved

• Model pickers now show which model Auto will actually use, even when another model is selected.
• Type / in Chat to find and insert available skills.
• You can now press Esc on the keyboard to stop a turn. If a skills popup is currently open Esc closes the popup first, then a second Esc stops a turn.
• Chat now holds its position more reliably while replies stream, older messages load, and you switch between conversations.
• Claude Opus 5 is now available and is the default for new agents.

New: Experimental Remote access

• Use Flashbag from another Tailscale-connected computer through password-protected browser access.
• Browser sessions sync messages across open clients, and can be revoked from the Mac.
• Remote access can only be managed directly in the Mac app. Some features (like the new attach files or folders from Finder) are not available via remote access yet.

Chat

• Attach files or folders from Finder to your next chat message, including queued messages.
• Queue one message while an agent replies, then edit it or stop the current reply to send it immediately.
• Sticky day separators and improved scrolling keep new replies in view while preserving your reading position.

Improved

• Choose layout density and activity-detail level in Display settings.
• Tasks now flag missing or inaccessible input sources and pause runs until they are repaired.

• Added GPT-5.6 Sol, Terra, and Luna plus Claude Sonnet 5, updated defaults and Luna or Sonnet 5 handling helper work based on which providers are available.
• Reasoning controls now show only levels supported by the selected model, add None and Ultra, warn when Ultra may rapidly increase usage through subagents, and default new agents to Extra High.

• Task times now use 24-hour local-time formatting consistently, weekly schedules respect the new “Week starts on” setting, and editing a task without changing its recurrence no longer shifts its next scheduled run.
• Agents now pick up task replies posted into their chat on the next continued turn, so they can respond with the same task output the user already sees.
• Agents can now inspect saved task runs through MCP, including run status, final responses, errors, output links, usage, and timestamps.
• Agents are now aware of how Flashbag stores dates and times, and your local time and can convert between them.

New: Tasks

• Create reusable automations for any agent, run them on demand, schedule them for later, and review every run from the new Tasks page.
• Schedule tasks by interval, weekday, or monthly day, with optional catch-up when Flashbag was closed during a run window.
• Watch task runs live, inspect the agent's work, open outputs, and send final results back into the agent's main chat when useful.
• Manage tasks from chat too: agents can create and maintain their own tasks, while Gordon can manage and diagnose tasks across agents.

Improved

• Context preload is clearer and more consistent across agents and tasks, with aligned message/fact timelines and better identity estimates.
• Chat and task output now render more cleanly, including headings, paragraphs, lists, tables, code, and Finder-opening file links.
• Dashboard refresh now recognizes Codex sessions with revoked refresh tokens as signed out.
• New agents default to Claude Opus 4.8 whenever Claude is available, with Codex helpers still used when available.

• Chat header token stats are more compact and no longer get cut off in narrow windows — each stat's breakdown now opens in a hover card (with a "?" hint) instead of crowding the header
• Chat messages and thinking steps now render Markdown
• Chat now warns when a conversation is nearing its automatic session refresh
• A failed turn now stays in its correct chronological place instead of jumping below newer messages after a reload
• Turn stats now show durations of a minute or more as "4m 15s" rather than raw seconds

• All-new chat turn view: thinking and tool steps stream in live as a flow strip, with per-turn stats and smoother scrolling
• Major tune-up of the built-in system prompts: agents follow their instructions more reliably, significant memory and fact improvements
• New token & cost tracking: a per-agent usage ledger feeds costs in the chat header (cache spend broken out) and in/out/cached breakdowns on every turn
• New dashboard usage chart: tokens or cost per day, filter by agent, per-segment hover values, agents keep their own colors
• Sturdier sessions: they survive app restarts, rotate to a fresh session before the context window overflows, and failures show the real CLI error
• Stopping a reply keeps everything produced so far, and the session stays alive so you can steer and continue
• Long chats stay fast: history is virtualized so big sessions no longer bog down the UI
• Gordon got a knowledge base (memory, skills, agents, session lifecycle) so his answers are more accurate
• Closing the window now hides the app — click the Dock icon to bring it right back (Cmd+Q still quits)
• Setup re-checks Claude/Codex sign-in on every launch and right after installs
• Sidebar busy dot clears after a stopped turn, and "last online" updates live while chatting
• Messages containing markdown headings now display correctly in transcripts
• Accurate token and cost figures on Codex: numbers no longer balloon as a conversation grows
• Token gauges no longer overcount turns that use tools
• The Bootstrap stat now shows where startup tokens go: identity, memory, and the CLI
• New agents default to GPT-5.5 (Opus 4.8 when only Claude is installed)
• Claude Fable 5 is available in the model picker
• Gordon greets once per day instead of on every visit
• "Started fresh" markers show up reliably, and model switches take effect immediately
• Chat page remembers which agent you were viewing (and your position if you scrolled up)
• Agents page remembers which agent you were viewing

• Gordon's reasoning effort capped at Extra High (Max tends to overthink on Claude)

• Claude Opus 4.8 is now available in the model picker and is the default for new agents
• New per-agent reasoning-effort control below the main model (Low → Max; new agents default to High)
• Levels a model can't honor map to its nearest available level, so switching models never breaks
• Gordon always runs at the highest available level

• Helper model now tracks the live role assignment instead of going stale mid-session
• Claude/Codex always run from ~/.flashbag/cwd so the working directory is predictable

• Light/dark theme switch now actually applies across the whole app
• Gordon's avatar is amber, matching his accent color
• Agent picker label is tighter
• Fixed a hang when reading your shell environment

• Faster startup: sign-in checks for Claude and Codex now run in parallel
• Codex works outside trusted Git folders again
• Agents launched by Flashbag now see the same PATH and tools as your own terminal

• Linux build
• Chat feels much better: streaming is smoother, new in-flight indicator, scroll no longer jumps
• New agents come with their helper CLIs already wired up
• Gordon picks his own model automatically
• Onboarding remembers what it found between restarts (no more redoing the same checks)

• Fixed a crash that could hit during import

• Gordon now handles agent identity during chat-log import end-to-end

New: Gordon

• Gordon arrives - a built-in assistant who sets up your agents, writes your skills, and answers questions about what your agents produced.
• Agents now have a type, so the assistant behaves differently from the agents you create yourself.

New: The designed interface

• The first designed interface replaces the bare one that existed only for testing.

New: Output folder

• Files your agents create now land in a dedicated output folder instead of being dropped in Downloads.

Improved

• Flashbag now runs on Claude alone. Until now it only worked if both Claude and Codex were installed.
• Skills are now linked correctly when an agent is created.
• The dashboard and the agents page show the real number of facts instead of zero.
• The activity dot now stays visible when an agent finishes instead of flashing and disappearing.
• Chat stops scrolling itself once you scroll by hand.
• Claude and Codex sign-in status refreshes when the app starts.
• The app no longer sends you straight to Gordon on launch.

New: More than one agent

• Create, edit, and delete agents in the app. Each one gets its own folder, its own memory, and its own skills.
• Agents run side by side - set one working, switch to another, and a dot tells you when each is done.

Improved

• The debug view was rebuilt: calls appear in the order they happened, and the internal searches that were previously invisible now show up with their own cost.
• The interactive prompts built into Claude are switched off, so a reply can no longer stall waiting on a question you were never shown.
• How long a reply may take before it gives up is now a setting.

• Fixed Codex failing on every call - it was being started without a flag it needs.
• Fixed Flashbag not finding Claude or Codex when they were installed somewhere other than the usual place.

New: The first build

• Chat with your agent from a desktop app instead of a terminal.
• Memory: your agent pulls facts out of the conversation as you talk and remembers them afterwards.
• Claude or Codex does the thinking, whichever you have installed.
• A debug panel on every message showing the tokens spent, the time taken, and each call that was made underneath.
• A dashboard that detects which command-line tools are installed and whether they are signed in.