Copera in Six Languages, Copera Loop & AI-Built Experiences

Copera now speaks six languages. Copera Loop turns a chat message into a live collaborative block anyone in the conversation can edit, Experiences lets you ask the AI to build an interactive page you can share, and the automation builder gains a real run debugger with run-now, retry, and step-by-step inspection.

New Features

German, French, and Italian

• Six languages, everywhere — German, French, and Italian join English, Spanish, and Portuguese across the whole product: the web app, notifications, emails, booking pages, the desktop app, public docs, and this documentation site. Dates, AI reply language, and email defaults all follow your new language too (#3416).

Copera Loop — Live Collaborative Blocks in Chat

• A document that lives inside the conversation — insert a Loop block into a chat message and everyone in the conversation edits it together, live, right in the message card. No opening a separate doc, no losing the thread (#3420).
• Templates to start from — insert a ready-made block from the composer toolbar or a slash command.
• Access that follows the conversation — a Loop block is readable by the people in the channel it was shared in; forwarding one warns you before it widens access, and creators or workspace admins can delete a block.

Copera Experiences — Interactive Pages the AI Builds

• Describe it and get a real, working page — ask Copera to build an interactive experience and it produces one you can open, iterate on, and version (#3436).
• Share it your way — keep an experience private, share it with your whole workspace, or grant specific people and teams viewer or editor access.
• Publish it — publish a finished experience so anyone with the link can use it.

Ask Copera — Docked Panel and Search Across Your Docs

• A docked copilot — Ask Copera now lives in a dockable side panel with its own rail, so you can keep it open beside whatever you're working on (#3431).
• It reads your docs — Copera builds a searchable index of your documents (respecting who can see what) and can search them, pull up an outline, or read a specific section when answering you.

Automation Builder — A Real Run Debugger

The automation engine got a deep overhaul (#3427):

• Run it now — trigger an automation from the builder toolbar, with a dialog that asks for whatever context the trigger needs.
• See exactly what happened — the run history is now a full debugger: per-step inputs, outputs, logs, and cost, status filters, retry from the run, and a click-through from a failed step straight to its node on the canvas.
• Test your webhook trigger — arm the trigger, send a real request, and the captured sample payload feeds the token pickers everywhere downstream.
• Sample data and health badges — board triggers offer sample data, nodes show a badge when required inputs are missing, and each node shows a one-line summary of how it's configured.
• Decide what happens on failure — every step can fail the run, continue past the error, or branch down a dedicated error path, with configurable retries and per-step timeouts.
• New control flow — an N-way switch, a stop block, sub-automations you can call from another automation, and a step that waits for an external callback.
• New transform blocks — aggregate, deduplicate, sort, filter, and slice arrays, plus CSV and JSON parsing.
• Schedules by the minute — agent jobs accept interval schedules from one minute to a week.

Workforce — Agents That Show Their Work

• A live activity timeline — when you DM or @mention a hired agent, it now runs a real multi-step, tool-using loop, and you watch it happen: each step appears as a verb → object → outcome row with an elapsed timer and a Stop button, then collapses into a receipt next to the agent's reply (#3418).
• Crash-safe turns — a turn survives a restart and resumes without repeating work or double-posting a reply.
• An answerable approval inbox — any action an agent needs sign-off for now appears in a "Needs your attention" inbox you can approve or reject in place, not just meeting actions (#3418).
• Agents that do the work from chat — agents can turn an image into a board row, write files, and tell you plainly which permission they're missing instead of deflecting (#3426).

AI Spend Controls

• Cap AI spend per member and per team — admins can set daily and monthly AI credit limits as a workspace default, per member, or as a shared team pool. A new Budgets tab on Settings → Billing shows where everything stands, and hitting a limit gives a clear dialog with the exact figures instead of a generic error. Everything is unlimited until you set a limit (#3425).

Bug Fixes

• Blank screen at startup — fixed a crash that could blank the entire app before the first page rendered, plus a root safety net that now shows a proper error card instead of a white page (#3434, #3435).
• The workspace dropdown that refreshed everything — opening the workspace switcher no longer reloads the whole app (#3433).
• Missing translations — 36 strings that were showing their raw key names (billing and member error messages, command palette errors, Inbox AI failures, Gantt undo/redo, microphone permission errors, poll voters, and more) are now translated in all six languages, and notification placeholders no longer leak into push and bell notifications (#3435).
• Getting logged out of everything — opening the workspace dropdown or leaving a single workspace no longer signs you out of every workspace and tab (#3435).
• Faster first load — the Ask Copera panel is no longer bundled into the initial download, cutting the entry payload by roughly 20% — even the login page was paying for it (#3435).
• Approvals button — the "Needs your attention" button now appears only when something actually needs your attention (#3437).
• AI credits — credits are only charged when an AI feature actually delivers a result (#3418).
• Chat — fixed blank channels and direct messages that could route without a counterpart, and the chat editor no longer resets when a Loop block is present.
• Audio playback — the player's play state and timer now stay in sync with the audio.
• Emoji — standalone emoji messages render larger and correctly.
• Voice transcriptions in search — meeting transcriptions that were missing from search have been reindexed.
• Screen sharing — fixed a crash when re-injecting a screen share.
• Website chat widget — the widget now retries its handshake for slow-loading pages, forwards the embedding site correctly, and recovers from a failed session instead of staying broken.

Improvements

• Mobile calendar stays fresh — the mobile apps are now woken silently to sync your agenda, so your calendar is up to date when you open it.
• Cleaner AI answers — stray reasoning fragments no longer leak into AI responses.