Product updates, use cases, technical write-ups, and company notes from the Conol team.
The Conol desktop app now works with an OpenWalkieTalkie deck: see your chats on the device, hold to talk to send a message or start a new chat, and answer the agent's questions without switching windows.
Read more →A small maintenance release with some bug fixes.
Read more →Streaming replies now save in small batches a few times a second instead of once every few words, so they put far less load on Conol and keep it fast and steady for everyone.
Read more →A spending history on the Billing page and a per-chat credit counter in the composer, four new models led by Claude Haiku 5.5 and Gemini 4 Argon, voice input restored, and a Sonnet 5.5 cache-billing fix.
Read more →Files and photos you attach while answering the agent now reach it, attachments show as clean file cards throughout the chat, and the spending history on the Billing page loads again.
Read more →Welcome credits and quest rewards now follow proof of a real person: they are paid when you verify your email (quests after your first day), sign-up recognises one mailbox behind address variants, and automated registration is harder.
Read more →A fix for a bug where attaching an SVG, HEIC or another image format the model can't look at made every later reply in that chat fail. Those files now reach the agent as attachments it can open and read.
Read more →A fix for a bug that kept files attached in chat from reaching the agent: documents, spreadsheets and other files now arrive with your message, and show as attachment cards in the chat.
Read more →Conol 1.1 makes every agent change to your notes reversible: version history, one-click undo from the chat, a Recently deleted shelf, and protection for edits you make while the agent works. Plus Claude Sonnet 5.5, GPT-6.1 Sol, a chat that stays connected, and a round of fixes.
Read more →Conol reaches 1.0: an agent that works inside your notes, keeps going in the background, reaches you wherever you are, and connects to the tools you already use. This release also brings a calmer view of the agent at work and a Mac sign-in fix.
Read more →GLM 5.3 FlashX joins the model menu as the fast option in the cheap tier, and signing in with Google on the iPhone app finally completes instead of dropping you on the home page.
Read more →A chat that needs your answer now says so, on its tab, in the chat list and in your inbox. Four new models join the menu, every capability tier gets a new lineup, and the app picks up a security update.
Read more →DeepSeek V4 Flash Vision can now read the images you attach, several models got cheaper after a full price re-check, and a handful of fixes land across the desktop app and the agent.
Read more →The main area becomes a two-pane tab workbench, the Mac app gets real multi-window support, every tab has a copyable identifier you can jump to with ⌘K, and CoHDL code is finally highlighted.
Read more →GPT-6 Astra in the model picker, and a fix for the worst thing an agent could do to your note: edit it successfully and then have the edit vanish. Plus edits that land more often, timers you can cancel from any chat, and a cheaper GPT-5.6 Sol.
Read more →A ⌘K command palette, keyboard shortcuts across the whole app with Esc to stop a running agent, full keyboard navigation for the sidebar, and GLM 5.3 Flash at a twentieth of the price. Plus a mascot that finally matches the logo.
Read more →Reorder your workspace list and have it stay put, pick up Z.ai's GLM 5.3, and let your agent reach a workspace you created mid-conversation. Plus the free-tier and premium-tool changes that come out of the bot-farming waves.
Read more →Attach files up to 200MB straight from the composer, jump back to the newest message in a long thread, and get a plain-language reason when a turn fails instead of a raw provider error.
Read more →What 12,000 fake accounts taught us about free credits — and why the Code tool and LLM egress are becoming premium-only.
Read more →Sign in to several accounts at once with each tab pinned to its own, bring your agent into a Slack channel, and a deep round of work on agent note edits that now leave everything they didn't touch alone.
Read more →Attio joins the connector list, generated media finally plays inside the iPhone app, agent reasoning streams live, and a long round of fixes for language, color, and non-Latin text.
Read more →How algebraic-effects-inspired control flow and durable mutations turn an LLM loop into a reliable runtime — and what a nine-hour stall taught us about where reliability actually comes from.
Read more →Anthropic's newest flagship model joins the picker, the model picker finally tells you the truth about your plan, and pasted images stop breaking.
Read more →The landing island becomes a playable hidden world, the Flash tier gets routed to genuinely fast hosts, task checklists learn to stay put, and files with non-Latin names finally deliver.
Read more →Pin the chats you keep coming back to, rename any session (and keep the agent from renaming it back), and let simple chats title themselves. Plus stronger sign-up protection and a notes-tree repair.
Read more →Model presets turn a long model list into four clear tiers with prices, the landing page gets a live voxel makeover, Settings becomes navigable, and a wave of fixes lands across web, channels, and iOS.
Read more →Conol joins your Slack workspace as a real bot: DM it like a teammate, let the whole team bind their own accounts, and get proactive updates where you already work.
Read more →Kimi K3 from Moonshot AI is now available in the model picker, at the same credit rate as our baseline.
Read more →Connectors that tell you when to reconnect, a clearer offline sign-in, and a round of chat and editor polish.
Read more →A personal trainer coaching 16 clients replaced scattered notes, texts, and memory with a Conol brief that connects scheduling to client context — cutting nightly prep from 45 minutes to 15.
Read more →Your agent can now run complex jobs as live, parallel workflows you watch as a graph; a new Effort control lets you set how hard a model thinks; you can add a profile picture; plus fixes for scheduled-session labels, duplicate names, and stuck sessions.
Read more →A K–8 school administrative coordinator replaced her morning routine of checking Gmail, Calendar, Telegram, voicemail, and shared sheets one by one with a single connected brief from Conol.
Read more →Watch the agent's shell commands stream their output in real time, enjoy dramatically smoother streaming in long conversations, and two fixes that make agent Q&A and tool use more dependable.
Read more →Workspace memory is now visible to every member of the workspace, scheduled plan changes say exactly which plan they move you to, and a round of small chat fixes.
Read more →A freelance operations consultant turned four hours of Friday client-update assembly into a 25-minute review, by delegating the gathering, summarizing, and drafting to a recurring Conol workflow.
Read more →OpenAI's brand-new GPT-5.6 family — Sol, Terra and Luna — lands in the model picker, and voice transcription moves to a new speech model that's markedly better at Chinese, dialects, and mixed-language speech.
Read more →A full-screen image viewer, resend for messages that don't go through, tighter read-only guardrails for workspace viewers, and a round of inbox, search, and rendering fixes.
Read more →A small skincare brand owner turned a repetitive Sunday competitor check into a Monday morning Conol digest, saving six to eight hours a month and freeing up time for customer research.
Read more →Conol now speaks Japanese, Simplified Chinese, and Traditional Chinese across the whole app and site; the inbox lets you read before it acts; the Gen tool makes video with Veo 3.1; agents draw on all of your memory; plus timer and composer polish.
Read more →After 4 months of solo building, Conol launches this week: an agent with a real note workspace built in. Your notes are its context, its work files itself, and it runs in the background.
Read more →New accounts start with 4,000 free credits, the blog is now a tap away from inside the app, and right-click copy & paste works in the macOS desktop app.
Read more →Reply to any message on Telegram and the agent keeps the thread, background tasks hear your answers, long conversations compact themselves, and your timers live in the composer.
Read more →A slim strip above the chat composer now shows live plan progress and a context-usage meter for long sessions, plus fixes for the Send button, collaboration cursor names, and workspace Skills & Memory permissions.
Read more →A fixes release: video and audio in publicly shared notes now play properly, agent replies on Telegram and WeChat link to real pages, and large generated or collected files upload reliably.
Read more →Anthropic's Claude Fable 5 has fully launched, and it's now available in Conol's model picker — a reasoning model with a 1M-token context window for long, involved sessions.
Read more →Claude Sonnet 5 becomes Conol's default model, the chat and sub-agent panels get resizable, and 0.36.0 adds read-only sharing, clearer connector errors, and a batch of editor fixes.
Read more →