Release notes

What changed in Conduit

A summary of each release, written in plain language rather than commit titles. The full list lives in the repository, and every binary is on the GitHub releases page. Conduit is in release-candidate stage for 1.0: installers are not OS code-signed, so Windows SmartScreen and macOS Gatekeeper warn on first launch.

1.0.0-rc.35 — 18 November 2026

Fixed: a workflow can keep its own browsing pace. Choosing one on a workflow's Start panel failed with “Unknown setting” in rc.34; it saves now.

Browser follow-ups. A workflow's run page says when the browser is waiting before opening a page, as a chat already does. A file the agent's browser downloads during a workflow run is also saved to downloads/ in the workflow's folder and listed with the run's results; it never replaces a file already there. Right-click a link in a chat reply or a source list to open it in the agent's browser, so you can sign in or pass a check there before the agent needs it (Settings → Computer must be on).

What's new after an update. The first launch after an update says so once, with a button that opens that release's notes here. Settings → Updates and Settings → About link to them too, and the GitHub release page now carries the notes.

1.0.0-rc.34 — 17 November 2026

The agent's browser browses at a human pace. By default it now pauses to read each page, opens at most six pages a minute on one site, and backs off when a site asks it to slow down, so sites are less likely to turn it away. This applies to everyone, including people who used the browser before. Waiting doesn't count against a turn's time limit, and when a page load waits a couple of seconds or more, the chat says so. Fast pace is still in Settings → Computer, and each chat (in its Browser menu) and each workflow (Browsing pace on its Start panel) can choose its own. Workflows that run while nobody is watching use the human pace unless they choose otherwise.

Fixed: a slide, draft or spreadsheet call with bad arguments no longer stalls a turn. The model now gets the reason, such as a missing field, and fixes the call on its next try.

1.0.0-rc.33 — 16 November 2026

Saved sign-ins. Settings → Computer → Saved sign-ins keeps a site, a username and a password; the password goes to your system's keychain. When the agent needs to sign in to that site, a card asks you first, then the app fills in the form itself. The model never gets the password: it can't type it, and wherever a page shows it, the model reads “(saved password hidden)”. If the saved sign-in doesn't get through, you're asked to finish. Workflows that run while nobody is watching can use a sign-in only if you turn on its switch, which is off by default.

1.0.0-rc.32 — 15 November 2026

Downloads ask first. When a click by the agent starts a download, a card shows the file's name and size and asks: Do it, Always on this site, or Don't. Programs and scripts are always refused, and so are files over 100 MB and more than five downloads in one run. An allowed file never goes to your Downloads folder: it is saved with the conversation, the model reads it, and a card under the step offers Save as….

The agent can look at the page. Models that accept images get a picture of the page in the agent's browser, for when the page's text isn't enough. The picture is saved with the conversation and sent to the model provider you chose.

1.0.0-rc.31 — 14 November 2026

The agent's browser refuses downloads. Before, a link or button that downloaded a file could save it to your Downloads folder without asking, and the agent wasn't told. Now every download is refused and the agent sees that nothing was saved. Asking before a download comes later. Site permission prompts (notifications, location, camera, microphone, clipboard reading, MIDI) are refused too, since nobody is watching that window. The Writing studio's side panel also stopped jumping back to Ask after you open another tab.

1.0.0-rc.30 — 13 November 2026

Clearer browser approvals in chat. The approval card says “first action on this site in this chat”, uses the normal interface font and no longer sits under a “Tool call complete” line while it waits for you. Browser steps show plain names such as “Opened page” and “Clicked”, element types on approval cards are translated in every language, and addresses show readable characters instead of percent-encoding.

1.0.0-rc.29 — 12 November 2026

Browser use. Workflow agents and helpers can read pages in your installed Edge, Chrome, Brave or Chromium, in a separate profile of its own, so pages that build themselves with JavaScript can be read. Local and private-network addresses are refused. Pick the browser, or reset its profile, in Settings → Computer. If a site shows a human check, you finish it in the browser window; Conduit never solves checks itself.

Click, type and choose. A workflow can let its agents act on pages. They ask the first time on each new site and before anything that spends money, changes an account or signs in. “Always on this site” and “Allow on all sites” remember your answer for that workflow. Typing passwords and your saved API keys is refused; when a site needs you signed in, you sign in yourself in the browser window and the agent never sees what you type.

Browse from a chat. A Browser button beside Web and Research lets the assistant use the browser in that chat, one message at a time, in the same tab. Approvals and questions appear in the chat and come back if the window reloads. Links in replies, sources and search results now open in your browser (before, clicking them did nothing), and links with brackets in them are no longer cut short.

1.0.0-rc.28 — 11 November 2026

Dictation. A microphone beside Send records you; click it again, or wait for the five-minute limit, and the text lands in the composer. Escape discards the recording. Your voice goes to the transcription provider you choose in the new Settings → Voice: OpenAI, OpenRouter or Groq, with your own key. Dictation is off in local-only mode. Until it's set up, the microphone has a dashed outline and opens Settings → Voice, and a setting hides the button.

1.0.0-rc.27 — 9 November 2026

Save a deck as PowerPoint. Slides' Export menu has a new option, “PowerPoint — slides you can edit”, on every platform. The file matches the deck as the app draws it: the same fonts, sizes, colours, cards, lines and backgrounds, in the same places. Text stays editable, charts are native PowerPoint charts (Edit Data opens their numbers), and speaker notes come along.

Read PowerPoint files. Attach a .pptx file to a message, or add one to Documents, and the model reads its slides in order, with tables, chart data and speaker notes. Macro-enabled .pptm files are still refused.

Read aloud. A speaker button under each reply reads it out with your system's voices, in the reply's language. Click it again to stop.

1.0.0-rc.26 — 6 November 2026

Charts in spreadsheets. Ask for a bar, line or pie chart and it appears under the sheet, in the app's colours for both themes. Charts follow the data when rows are added or removed, and the downloaded .xlsx has real Excel charts.

Save as Word. Writing documents export as a Word file (.docx) next to Markdown and HTML, with real headings, lists, tables, links and code blocks.

Approvals say what will happen. Each approval opens with one plain sentence, such as “Run ‘npm test’ in ferry” or “Write report.md (2.1 KB) in Documents”; the tool, arguments and folder are still shown below. The assistant can also rename a spreadsheet.

1.0.0-rc.25 — 6 November 2026

The window no longer opens black on older macOS. On macOS Monterey the interface couldn't be read by the system WebKit and nothing drew. It is now built for Safari 14, which covers macOS 11 and up.

1.0.0-rc.24 — 6 November 2026

Ollama works after tool calls again. The next request after any tool call was rejected with a “can't find closing '}'” error because earlier tool arguments were sent as text; they now go as an object. Once per launch Conduit also checks that its window actually appeared, and if it stays blank it writes a boot-check.txt report to the log folder and says where to send it.

1.0.0-rc.23 — 5 November 2026

Opening a chat brings back its latest artifact. The panel opens the artifact you last had open, or the most recently changed one, instead of the list. A chat you left on the list still opens on the list. Every macOS build is now launched before a release is published.

1.0.0-rc.22 — 3 November 2026

Conduit starts on macOS again. It crashed on launch on every Mac with “a webview with label main already exists”; the window also opens at its proper size again.

Spreadsheet fixes. Euro, yen or yuan sheets no longer show $, and formats such as R$ or CHF no longer turn cells into #VALUE!. An inserted row takes the formats and formulas of the row above it, and a total that ends just above it grows to include it. Writing into a formatted cell keeps its format, numbers sent as text are stored as numbers, and fewer edits fail on small differences in a model's wording. The panel shows numbers the way your app language writes them. A request for a “Tabelle” or a budget mentioning a version number now builds a spreadsheet instead of a web page.

1.0.0-rc.21 — 2 November 2026

Spreadsheets. Ask for a budget, a price list or a loan calculator and the chat builds a real spreadsheet: formulas that recalculate, currency, percent and date formats, several sheets, a frozen header row. It opens beside the chat, and you change it by asking (“add a row for March”, “make column C euros”). Export it as an .xlsx file that opens in Excel or Google Sheets. Charts are not in spreadsheets yet.

Excel files as attachments. Attach an .xlsx file to a chat, or add one to a collection under Documents, and the model reads its sheets as tables.

You can see when a model has finished editing. Under the reply, a line says “✓ Done · Updated “Menu” · 3 changes”, or “Still working on…” while it runs; click the title to open the document. The document panel shows “Editing…” and then “✓ Done”. A chat that finishes while you are in another one shows “Done” in the sidebar, and if Conduit is in the background when a long reply finishes you get a system notification (Settings → Agent turns it off). A model editing a document in several passes now finishes them all; before, the turn ended after the first edit.

An update notice that stays. With update checks on, a new version shows a message that stays until you close it, with Install and restart; with automatic updates, a downloaded update offers Restart now. Conduit checks each time it starts, and an install you start during a reply waits for the reply to finish.

Smaller changes. Document titles a model wrote HTML-escaped (“Maple & Rye”) now read “Maple & Rye”. The document panel no longer shows “(inline payload)” for a stored document. Model prices were refreshed from models.dev, and dependencies were updated.

1.0.0-rc.20 — 1 November 2026

Six new providers. Cerebras, SambaNova, DeepInfra, Novita AI, MiniMax and NVIDIA NIM. Add a key in Settings → Providers and their models appear in the model picker, each with a base URL field. SambaNova, DeepInfra and Novita AI list their prices, so their models show what they cost. Conduit now supports twenty-three providers.

Fewer wrongly blocked tool results, and a connector you name is used straight away. A web page or document that merely contained a phrase like “act as a …” or “you are now …” in the middle of a sentence was stopped as a prompt-injection attempt. Now only text that opens a sentence with an instruction counts, and your own answers to an agent's question are never blocked. When your message names a connector (an MCP server), its tools are available at once instead of being hidden behind tool search.

Smaller changes. Fields in the agent editor, project settings, workflow editor, branding, the agent's question box and app details read their value as you type, so a fast edit can't come out blank. The “no artifact content” notice no longer appears after a turn that started a slide deck. Model prices were refreshed from models.dev, and dependencies were updated.

1.0.0-rc.19 — 31 October 2026

Retry keeps every answer. Retrying the last reply no longer throws the old one away. Each retry is kept as a version, and a ‹ 2 / 3 › switcher under the reply moves between them; the chat continues from the one showing. An earlier reply shows its versions but cannot switch, because the messages after it were written from the one it has.

Retry with another model. The arrow next to Retry opens the model menu. The model you pick answers the last question again, for that reply only: the chat keeps its own model. The command palette has “Retry with another model…” too.

Compare. Compare, next to the switcher, opens the versions side by side in the panel on the right, each with the model that wrote it, its answer and its cost. “Use this one” picks a version for the last reply, and “Fork from this version” starts a new chat that continues from any version. A reply now records the model that wrote it, so its model line still names the right model after the chat is reopened.

Going back a version. This release adds a database migration. If you go back to rc.18 afterwards, it says your data was opened by a newer version and leaves it alone; install rc.19 or later again to open it. rc.17 or earlier sets the database aside as conduit.sqlite.corrupt-<time>.bak in your data folder and starts empty; your data is still in that file. Model prices were refreshed from models.dev.

1.0.0-rc.18 — 30 October 2026

Projects. The rail's Folders are now Projects. Any project can have settings that every chat in it uses: instructions, Documents collections searched in each chat, skills turned on in each chat, and a working folder. Open them from the project's ⋯ menu → Project settings. A project with no settings is just a folder, as before.

Instructions stack. Your Settings instructions come first, then the project's, then the chat's own, each labelled for the model. A chat's own collections and skills are added to the project's. Settings are read when a message is sent, so editing a project, or moving a chat in or out of one, applies to its next message.

New chat here. In a project's ⋯ menu it starts a chat inside the project, in the project's working folder, so the model can read and write files there and the folder's AGENTS.md goes with every message. Chats already in the project keep their own folder, and any chat can still change it. In a chat, a project chip in the composer opens the project's settings, and the chat settings popover lists what the chat also uses from Settings and the project. A project's ⋯ menu also has Rename and Delete; deleting asks first and keeps the chats.

Going back to an older version no longer empties your chats. A version that finds your data was last opened by a newer one now says so and leaves the data alone, instead of setting the database aside and starting empty. This protects going back to this release or later. This release adds a database migration: rc.17 or earlier cannot open the data, sets the database aside as conduit.sqlite.corrupt-<time>.bak in your data folder and starts empty. Your data is still in that file; reinstall rc.18 or later and put it back. The composer's popovers (chat settings, skills, documents, connector prompts and resources) now open downward when there is no room above, so in a new, empty chat their top is no longer cut off.

1.0.0-rc.17 — 28 October 2026

Continue a cut-off reply. When a reply stops at the output limit, the note under it has a Continue button. The model picks up exactly where it stopped, in the same message, with no new message from you. It works with every provider.

Skills from a chat, or from GitHub. The command palette's “Make a skill from this chat” has the current model draft a skill's name, description and instructions from what you asked. You edit the draft; nothing is saved until you choose Save, which also turns the skill on for that chat. Tool output and document text are never sent to the draft. Settings → Skills also takes a github.com/owner/repo link or a link to a folder in it. Only that folder is unpacked, and scripts in a skill are never run. A name that already exists now offers to replace the skill instead of failing.

Quick ask shortcut. Off by default. Set one in Settings → Chat and it opens a small ask box from anywhere; Enter sends the question to the app, which opens it like the Home ask box. Esc or clicking away hides the box. It is not available on Linux Wayland, where global shortcuts generally don't work.

Tool search for many connectors. When a chat would offer more than 20 connector tools, or their descriptions run long, the model gets a search_tools tool instead of the whole list and loads the tools it finds for that turn. With fewer tools nothing changes, and approvals work as before.

Smaller changes. On Windows, a connector started from a command, such as node or npx, no longer opens a console window that reappears when you close it. A new release now stays a draft until every installer is attached, so download links no longer fail while a release is being published. Requests to OpenRouter use its current attribution header names. Model prices were refreshed from models.dev. pulldown-cmark was raised to 0.13.5 to address exponential event growth with ENABLE_WIKILINKS, which Conduit does not turn on.

1.0.0-rc.16 — 26 October 2026

HTML files in chat and Documents. HTML files can be attached to a chat and added to Documents, alongside PDF, Word, text, Markdown and CSV. The page's title and text are read; scripts, styles and other code are left out.

A tidier Settings. Usage & cost and Updates have their own sections. About shows the version, links to the website, source code and issue tracker, the license, and where the app keeps its files on this machine. The command palette opens the new sections directly. undici, used only by the test suite, was raised to 7.30.1 for three advisories; the app itself does not ship undici.

1.0.0-rc.15 — 26 October 2026

Mark a connector tool “Only reads”. In Context → Connectors, each tool of a running connector now shows what it declares: only reads, changes things, or always asks. A tool that declares nothing asks before every call; tick Only reads for one you trust to only look things up, and it runs without asking in chats, agents and workflows. A tool that says it changes things, or is sensitive, keeps asking. The approval card for a tool that declares nothing now says so, instead of “may change an external system”, and points to Only reads.

Agents, a little clearer. A workflow step waiting for you shows under “Needs you” on the Agents page and reads “Waiting for you” in the run view; it no longer counts as a running agent. A chat asked to one of your agents shows that agent and its model in the status line. An agent used by a routine says so on its card. An agent that remembers is told to save what its next run needs when a task compares with earlier runs. Reading web pages now handles compressed pages (gzip, Brotli, deflate), and a page that blocks automated readers says so instead of returning unreadable text.

Fixes. The model picker in a new chat had no search field because the top of the menu was cut off. A tool result over 50,000 bytes with non-English text could crash the reply. A reply that stopped on an internal error left its helpers “Working” forever; it now ends them and says the reply stopped. Ordinary review pages containing text like “Assistant:” mid-line were blocked as prompt injection. A redirect sent to a helper showed after its answer and was gone after a restart; it now shows where it was sent and is kept. Text typed for a helper went to the main chat if the helper finished first; it is kept, with a note that the helper finished. Clicking a workflow step under “Running now” now opens the run, Stop no longer scrolls out of sight in a long agent view, and a toast no longer announces a helper finishing while you are watching it. The Workflows list stopped saying “Waiting for you” after you answered, saving one agent setting no longer resets the others, and “Save it as a document” in a chat is recognised as asking for a document.

1.0.0-rc.14 — 24 October 2026

Your agents. On the Agents page you make an agent once (name, instructions, model, tools, the agents it may hand work to, and a token limit) and use it in any workflow. A workflow keeps a reference, so editing the agent changes it everywhere. Four starters come with a new profile: Researcher, Writer, Reviewer and Data analyst. Deleting an agent gives each workflow that used it its own copy. Workflow helpers can come from your agents (Add from your agents), be copied into one workflow (Make a copy here), or be saved to your agents (Save to your agents).

Agents in chat. Type @ in the message box to pick one of your agents for that turn, or press Ask on its card. In Settings → Agent you choose which agents the assistant may hand work to; each hand-off runs on that agent's own model and shows as one line in the reply, with its status, tokens, cost and time. Several hand-offs in one go run at the same time (up to 3 by default, 1–4 in Settings → Agent); helpers that change things, such as using your computer or a connector action, still work one at a time. Under a reply, Helpers totals tokens and cost by model.

Open an agent to watch it. Clicking a hand-off shows that agent's live view: its task, tool calls, tokens and cost. Stop ends that helper only, and Redirect (or the message box) sends it a new instruction while it works. A toast says when a helper finishes or needs approval, the Agents rail button counts running agents, and the Agents page lists what is running now. A workflow run shows its agents as a tree under each step, with totals by model and the same Stop and Redirect controls.

Routines. “Run on a schedule…” on an agent's card gives it a task to do daily, on weekdays, every few hours, or when a feed or folder has something new. Each answer lands in Results, with an optional notification. A reply from one of your agents offers “Schedule this…”, and “Turn into a workflow” makes a routine an ordinary workflow you can edit.

Agents that remember and read, and a review on request. An agent set to Remembers keeps its own notes between runs, which you can see and edit in its editor. Reads lets it search the knowledge collections you pick. The document panel's ⋯ menu can send an HTML, Markdown or text document in a chat to your Reviewer agent; if it finds problems, the chat revises the same document, otherwise a toast says it looks fine.

Smaller changes. The workflow side panel can be dragged wider or narrower and keeps its width. A chat turn stops after 15 minutes; a helper's waiting time doesn't count against the assistant's clock. A helper's tokens were counted twice, and a helper that reached its token limit lost its work; it is now warned at 75% to answer with what it has, and a finished answer is kept. Two helpers in one workflow step that both needed approval no longer overwrite each other's request, and stopping one helper no longer cancels other turns' approvals or questions. Dependencies were updated for security advisories, including rustls-webpki (the distrusted emSign intermediates are no longer accepted), tokio, vite, katex and dompurify. A profile opened with this version can't be opened by an older build.

1.0.0-rc.13 — 20 October 2026

Phone and tablet layout. Below 720px wide the rail becomes a bottom tab bar (Home, Chats, Apps, Inbox, More), with the other areas and the theme switch in a More sheet. The chat list and the document panel open full width, two-column pages (Workflows, Settings, Context, the Slides and Writing studios) stack, and on touch screens hover-only actions stay visible with larger tap targets. The window can now be as narrow as 360px. Below 820px of height the rail turns compact and its destinations scroll instead of squeezing together, with Settings kept at the bottom and the light/dark switch moved to the title bar. Resizing the window no longer lags, because animations pause while you drag.

More of the app in your language. Research plans, findings and reports follow the language of the question, and the report's own headings and notes (Summary, Findings, Sources and the rest) come in all eight languages. So do the agent's progress line (“Writing the next section…”), the reason a computer-use card asks, the buttons on an unopened document card, and starter apps' names after you switch languages. Rail labels shrink to fit instead of breaking inside a word in German, Spanish, French and Portuguese.

Reload-proof replies, and one command outside the folder. If the window reloads while a reply is still working, or waiting on one of its questions, the chat finds it again: Stop comes back, and so does the question's form. On Windows, a build tool such as cargo can fail inside the folder limits; when you are watching the run, a second card offers to run exactly that command once outside the folder. Nothing is remembered, and unattended runs only tell the agent why it failed.

Fixes. Replies no longer start with “[App record of the tools you called…]”; that record is kept out of the model's view of earlier replies, and old copies are removed from history. A question a reply is waiting on comes back when you leave the chat and return. After a restart, tool rows and “Thought for…” show their real durations. In Writing, a model that answers the “write the draft” step without writing anything is asked once more for the first section instead of leaving an empty draft. A page from the chat you just left no longer opens in the next chat's panel, so Save as app can't save the wrong page; sites listed with commas are split into separate hosts, and long lists wrap. A page that fails to load a script now names the script (and its bad integrity hash, if any), so Ask to fix can fix the real cause. The artifact panel opens wide enough to use on a narrow window, an app you just opened no longer reads “Not opened yet”, and an open app's ⋮ menu opens under its button. The Agents page's working status stays a tidy box of up to three lines, with the full task in its tooltip.

1.0.0-rc.12 — 15 October 2026

Workflows you can see and build on a canvas. An open workflow shows its steps as a canvas or as a list, and Conduit remembers your choice. The canvas draws the start, each step, loops and the lines between them. During a run each step shows whether it is running, finished, failed, skipped or waiting for you, and a loop counts its items. In the editor you add steps from a palette, drag them to reorder or into a loop, and drag a wire from one step to another to choose which output goes into which field. Steps with problems are marked, and leaving with unsaved changes asks first. An open workflow has Steps, Runs and Results tabs, with the canvas filling the page.

Branches, steps side by side, and joins. A new “If… otherwise” step sends the run down a Yes or No path, and the path not taken shows as “Not taken”. A step can wait for any earlier steps, so steps that don't depend on each other run side by side, up to four at once. A “Runs after” control on each step does the same without dragging. Workflows saved before this run as they did.

Results you can find. Every finished run keeps what it saved, with a preview. The workflow page opens on its latest result, a Results tab lists them, and a Results list in the Workflows sidebar covers every workflow, with a dot on the ones you haven't seen and “Mark all as read”. Home shows “Fresh from your workflows”, and a scheduled run's notification offers Open. The Morning briefing starter now reads eight stories that match what you care about, follows each link, and saves a new dated briefing every run, grouped by theme.

Agents that ask before they act. When a step wants to use a connector tool that sends, posts or edits something, the run pauses and shows the exact call. Choose Send, Edit and then send, or Don't send, or allow that action from now on in this workflow. “Outside actions” in the Start panel turns asking on or off per workflow and lists what is allowed; tools a connector marks as sensitive always ask. Agent steps can use up to ten connector tools. A new Inbox collects everything your workflows are waiting on, with a dot on the rail and a system notification for new requests.

Helpers and an Agents page. A workflow can define helpers, each with its own instructions, model, tools and token limit. An agent step hands tasks to them and gets their answers back. A helper can hand work on one more level at most, and a step hands out at most twelve tasks. The Agents page shows a card for each agent step and helper with whether it is working, its model, tools, hand-offs, and tasks, failures and tokens over the last 30 days.

Agents that use your computer, after you approve. Off until you turn it on in Settings → Computer. A workflow's agent step or helper can then run commands and read, search, write or edit files in one folder you choose. Every command asks first unless one of the workflow's allow rules covers it. Some commands always ask, such as deleting files, git reset --hard, force pushes, system tools, shells, and long or unusual commands. After the agent reads outside content, such as a web page or a connector result, allow rules stop applying for the rest of that run. File changes show a diff and ask first, unless you let the workflow change files freely in its folder. Risky commands get their own card with Don't run as the main button, and “Run and allow…” shows the rule it would add. A Permissions tab on each workflow lists its command rules and how often each was used, the built-in always-ask list, and rules suggested from commands you approved in the last ten runs. Run details list every command and file change, who approved it, the exit code and the output. On Windows, commands are also kept from writing outside the folder; macOS and Linux have no such confinement, so your approvals are the control.

Runs that wait now survive a restart. A run paused on a question, a review or an outside action carries on after you close and reopen Conduit, with its original deadline. Steps that finished don't run again and an action already sent isn't sent twice. A run whose workflow changed while it waited stops and says so. Time spent waiting for you no longer counts against a run's time limit.

More ways to start, and drafting that knows them. A workflow can start when a connector's read-only list tool returns something new, checked every 15 minutes to once a day, and each new item starts a run. Describing a workflow can now draft branches, steps side by side, connector steps, helpers and the computer tool. Anything the model invents, such as a connector you don't have, is removed and noted.

A grouped rail, an Apps home screen and Context. The rail is grouped: Home; Chats, Slides, Writing; Apps, Manage; Workflows, Inbox, Agents; Context; Settings at the bottom. The place you are in shows a solid icon. Apps is now a page of cards with each app's name, category, when you last opened it, what it does, and whether it uses the internet or your model, with search and a New app button; Manage keeps the previous Apps page. Documents, Library, Connectors and Memory are tabs in Context, which reopens on the last tab you used. Ctrl+1 to Ctrl+9 now open Home, Chats, Slides, Writing, Apps, Workflows, Inbox, Agents and Context.

Examples for a new profile. A new profile starts with a World weather app, the chat that built it, a plan for a rainy Saturday, an 8-slide deck on how LLMs work and a Morning briefing with a result waiting, in your language and with no API key. Each is marked Example, and Settings → Privacy & data → Remove all examples deletes them.

Easier to read. No text in the app is smaller than 12px. Dark has brighter secondary text and clearer cards. Light has a cooler background, white cards, and stronger lines and text. Labels are in sentence case, status chips are clearer, and disabled primary buttons are grey. There is one accent colour across the app; Slides, Writing, Apps and Workflows keep their own colours on Home only. Each page title sits in a bordered band and leads with its rail icon.

Smaller changes. The workflow canvas no longer stays on “Running…” after a run ends. Choosing Don't send on an outside action stops the run with a reason instead of failing it. A branch that answers “no” no longer makes the run read as “Nothing new”. Links in a workflow's document ask first, then open in your browser. Markdown shows links and code inside bold and italic text. Requests to OpenRouter are now credited to conduitllm.com. Conduit updated rustls to 0.23.46, which addresses four security advisories. A profile opened with this version can't be opened by an older build.

1.0.0-rc.10 — 8 October 2026

Workflows that start themselves. A workflow can start when a new post appears in a feed (RSS or Atom, checked every 15 minutes to once a day) or when a new file lands in its folder. Turn it on with Run automatically: the first check only notes what is already there, then each new post or file gets its own run, with its title, link or file path ready for the steps. A file still being written waits until it's finished; hidden and temporary files are ignored. If the feed or folder can't be read three times in a row, you get one notification and the workflow says why it's paused; it carries on by itself once it can read again. Starters: “Summarize new posts” and “Inbox folder”.

Describe a workflow and Conduit drafts it. Say what you want in your own words — on the Workflows page, or with the Automate chip on Home — and the chat model drafts a workflow, checks it, and fixes what doesn't check out. It opens in the editor unsaved, with a short list of what's left for you to pick (the folder, the deck); nothing is saved or scheduled until you say so. Save chat as workflow in the command palette turns what you did in a chat into a workflow you can run again; it sees your messages and which tools were used, not the pages or documents themselves.

A model per workflow, and per step. A workflow can use its own model instead of the one picked in chat, and each summarise or agent step can override it: a fast, cheap model for the bulk of the work and a stronger one for the final write-up, or a local model for steps that touch private data. Workflows with no model chosen keep following the chat model, and each step's run shows the model it used. If a chosen provider isn't set up, the step uses the chat model and says so; in local-only mode a step set to a cloud provider stops with a plain message. A scheduled workflow asks again before sending to a provider it hasn't been approved for.

Only speak up when something changed. A new “Continue only if…” step checks a value and ends the run quietly when the check fails: has it changed since the last run, is it empty, does it contain or equal some text. A scheduled run that finds nothing new sends no notification and shows “Nothing new” in the run list. Notify and save steps can also skip themselves when what they would send or save is the same as last time. A new “Watch a page” starter puts it together.

Workflows read your data and your work. A workflow can have its own folder, and a “Read a file” step reads text, CSV, JSON, Markdown, PDF or Word files from it, never from anywhere else; a scheduled run asks before reading a folder it hasn't been approved for. “Turn text into a table” reads CSV, TSV or JSON into rows and columns that later steps can use or loop over. “Update a deck” and “Update a draft” steps change a saved Slides deck or Writing draft from data or instructions: new numbers in a chart, a new “This week” section. Text you wrote yourself is never changed, the edit runs in the document's own chat (marked as coming from the workflow), and a version is kept before and after so “Undo this update” in the run, or History, puts it back. Starters: “Weekly numbers deck” and “Monthly report section”.

Research, Documents search and connectors in workflows. A Research step runs the same research as in chat and saves the report, at quick or standard depth and on the step's own model; a Documents search step finds the passages in your collections that answer a question, with citations, for later steps to use. A “Use a connector tool” step calls a tool from one of your connectors — list a folder, read a file, list a repository's open issues. For now only tools that just read can run: anything that could change something is shown greyed out and refused when run, without asking. Scheduled runs ask before researching the web, searching your collections, or using a connector they haven't used before, and a Documents search never asks to send your documents for embedding on its own — it says so and stops. Starters: “Weekly research digest” and “Check notes against my docs”.

Export a file, or suggest a memory. An “Export a file” step saves Markdown, text, CSV, JSON or HTML into the exports folder, with Show in folder on the run. A “Save to memory” step suggests a memory for you to accept; it never adds one on its own.

Fixes. Connectors that only offer tools, like the official filesystem server, no longer show “0 tools”, and npx works as a local connector's command on Windows. Research reports no longer show internal labels like “[C51–C53]” when a range of facts is cited. A dead site no longer breaks a workflow: a step set to skip on error now lets the run go on, and the error says which step was skipped and why. Answer and Review on Home open the workflow that needs you; leaving the Workflows page during a run and coming back keeps Run now off and Stop available; an answer over 2,000 characters keeps your text, with a counter. When a step reads several pages, each gets a fair share of what the model is sent (150,000 characters in all). run.date and run.time are your local date and time, not UTC, and workflow errors are clearer, naming a misspelled setting or the line and column of a JSON mistake.

1.0.0-rc.9 — 7 October 2026

Slides built from their parts. The assistant no longer writes a slide's HTML: it picks a layout and fills in its parts (a headline, points, numbers with their labels, columns, a quote) and the app lays the slide out in the theme. Text too long for its place goes back to the assistant with what to shorten before the slide is made, and a free-form layout remains for slides no layout fits. Bar, line and funnel charts are drawn by the app from the numbers the assistant gives, with axes that start at zero and labels that don't collide. After each change the app looks at how the slides render: if text runs off, overlaps or crowds the footnote, or a diagram is unreadable, the assistant gets one round to fix exactly that, and anything left is shown to you as a note. Text you wrote yourself is never changed by it. The check catches those problems; it doesn't make a slide well designed.

Research rates its sources and checks its report. Each page it reads is rated for credibility. A weak source is marked in the report's sources, and its claims are attributed to it or left out, never stated as established fact; official, public-body and university sites are never rated low. The rating is the model's judgement; the quote check, done by the app, is unchanged. A review pass then checks every sentence against the facts it cites and corrects or removes what they don't support. Research also knows today's date, explains figures that differ by year, edition, definition or units instead of calling them a disagreement, and an unanswered sub-question names the sites that couldn't be read.

Fixes. Usage & Cost now counts every model call: Research runs, workflow steps and chat turns without tools were missing from it and from the daily spend alert. With OpenAI-compatible providers (OpenAI, OpenRouter, DeepSeek, Groq and others), several tool calls in one reply now run in the order the model wrote them, so slides no longer land out of order. A document you sent shows as a file chip with its name and type instead of a broken image, and the composer's + menu now says Attach files….

1.0.0-rc.8 — 6 October 2026

Attach documents to a message. Drop a PDF, Word, text, Markdown or CSV file on a chat, paste it, or pick it with +, and the model reads it. Every model gets the document's text, extracted on your machine; models that read PDFs themselves (Claude, OpenAI, Gemini, and OpenRouter models that list file input) get the PDF too, so charts and scanned pages come through. Either way the content goes to the chat provider you picked. Each attachment says how it will be sent, or why it won't be. In Chats a drop anywhere on the chat attaches; elsewhere a drop still adds to Documents. The assistant can also read PDFs and Word files in an attached workspace folder.

Faster to choose, easier to watch. The model menu opens with a search field, and each provider's models are listed alphabetically. A new Reasoning effort setting (Auto, Low, Medium or High) trades thinking for speed and cost; it is sent only to models that accept it, for OpenAI, OpenRouter, Anthropic and Gemini. Settings → Usage & Cost can alert you once a day when today's estimated spend passes an amount you set; it is off by default and never stops anything. Requests to Claude, directly or through OpenRouter and OpenCode Zen, now use Anthropic's prompt cache, and cache reads and writes are priced at their own rates.

Slides that fit. Slide text shrinks a step at a time, down to a readable size, instead of overlapping or running off the slide, and exported HTML and PDF match what you see. A full-width chart layout gives a chart the whole slide. Big decks now finish building instead of being cut off partway. Fixes: DeepSeek V4 Flash and its vision variant now get attached images (text-only DeepSeek models stay text-only), and the context gauge takes each model's window from the bundled catalog, so DeepSeek V4 shows 1M instead of 128K.

1.0.0-rc.7 — 5 October 2026

Writing: long pieces, outline first. A new area on the rail for blog posts, technical docs, reports and newsletters. Describe what you're writing and the assistant proposes an outline, with the sections, what each must say and a target length, that you edit and approve. Then it writes the draft section by section into an editor beside the chat. It can draw on the web (off by default), your document collections or a finished Research report, linking each fact to its source and marking anything it can't source as a TODO instead of making it up. Ask for a change and only that part is rewritten, or select text to make it shorter or clearer. Text you write yourself stays word for word, History keeps every version, and a draft exports as Markdown or HTML.

Costs for every current model. Prices now come from a snapshot of the open models.dev catalog bundled with each release, about 790 models, instead of a short hand-kept table; OpenRouter models use the price OpenRouter lists. You can set the price of any model yourself, and a model with no known price says so instead of showing $0. Costs were also shown 100 times too small, and cached input was counted twice on several providers; both are fixed, and past usage is re-priced.

Fixes. DeepSeek chats no longer fail after the first reply. Time spent on a tool approval no longer counts toward a turn's time limit. Asking to fix a page updates that page instead of writing a new one. Hosted web search can now read the pages it finds, and a later reply no longer mistakes earlier search results for made-up ones. The inspector's Sources tab lists a Research run's sources, and Home's Start a deck chip starts a deck from what you typed.

1.0.0-rc.6 — 3 October 2026

Research: a report whose every finding is checked against its source. Turn on Research in the composer's + menu (or pick it on Home) for a question that needs more than a quick search. It proposes a brief, with the sub-questions it will answer, a scope and a depth (Quick, Standard or Deep), and nothing runs until you approve it. Then it searches, reads the pages in full and pulls out facts, each with the exact words from the page behind it. A fact counts only if those words are really on the page, and the app checks that, not the model. The result is a report in your chat's documents: a summary, findings per sub-question, open questions and numbered sources, with every finding cited. Stop keeps what was found so far. Research needs web search turned on and isn't available in local-only mode. Page text only reaches a model call that has no tools, so a page that tries to give instructions can't make it search, fetch or save anything.

PDFs on the web are read. Research, a workflow's Fetch page step and the model's page fetch now read PDFs up to 20 MB, parsed on your device the way Documents imports one. A scanned PDF with no text counts as a page with nothing to read.

Security: connector tools ask unless they're marked read-only. A tool from an MCP connector now runs without asking only when its server marks it read-only (the standard readOnlyHint, or Conduit's own field); every other tool asks first. Before this, tools on most third-party servers ran without asking, because only a field that Conduit alone reads could make one ask. Connectors you already use may now ask for tools that ran silently before, and "Always" on the prompt remembers your answer, as before.

Security: the model's page fetch can't reach your own network. It could previously be pointed at localhost or a private address. It now fetches public https sites only, re-checks every redirect, and returns the page's readable text instead of raw HTML.

A search-heavy answer no longer stops silently. Past a turn's search limit, a chat used to end on whatever the model had said so far, with no answer and no error. Now the model is told to answer with what it has. With Exa, Tavily, Brave or SearXNG a turn can search up to 8 times (DuckDuckGo stays at 3), and those results now show as sources in Activity and Sources.

1.0.0-rc.5 — 3 October 2026

Web search on every model, with no setup. Search now works out of the box on any model. OpenRouter uses its own built-in search, billed to the same key, like OpenAI, Gemini and Anthropic already did. Every other model — Groq, DeepSeek, a local Ollama model — searches through Exa by default, which returns real web results with no key (free and rate-limited; queries go to Exa under its terms). DuckDuckGo's instant answers, the old default, rarely found anything; if you chose it, it stays selected, and Settings → Web search offers a one-click switch.

Get a search key in one click. Settings → Web search links to the key pages for Exa, Tavily and Brave, and to SearXNG's setup guide, each opening in your browser.

Personal apps. Saved apps are now called personal apps on the Apps page and on Home; the rail still says Apps.

1.0.0-rc.4 — 2 October 2026

Home. Conduit now opens on a Home page. One box starts anything: ask for a deck and Slides opens with a storyline, ask anything else and a chat starts. Below it, Home shows what is waiting on you (workflow approvals and questions, memory suggestions), your latest chats, decks and apps to pick up, and every part of the app as a tile with a live count. A part you haven't used yet shows an example to try instead.

Go anywhere from the keyboard. Ctrl+1 to Ctrl+9 (⌘ on macOS) open Home, Chats, Apps, Slides, Documents, Library, Workflows, Connectors and Memory, and the command palette has a “Go to” entry for each.

Fixes. Remote connectors on the 2025 protocol, such as DeepWiki, connect again instead of showing as down. A chat opened from Home starts at its latest message. The Pitch deck idea now opens in Slides. A connector search that hits a network hiccup tries once more before it gives up.

1.0.0-rc.3 — 2 October 2026

Slides. A new Slides section for presentations you build by asking. Describe the story, or ask for a deck in any chat, and the assistant drafts a storyline you edit and approve before it builds the slides. Everything happens in one live deck, so a change edits the slide it's about instead of producing a new copy of the file.

Edit it your way. Ask for a change and the assistant edits one slide at a time, or swaps a word across the whole deck. Script shows every word of the deck as one document you can type in, and you can double-click text on a slide to change it. Text you write stays as you wrote it when the assistant reworks a slide, unless you ask it to change that text. History keeps a version for every change, named after your request.

Present and share. Present full screen, with a separate presenter view (current and next slide, speaker notes and a timer) you can keep to yourself while sharing the slides. Export a deck as a single HTML file that presents itself in any browser, or as a PDF on Windows.

Fixes. The Documents page, the indexing consent dialog and the usage table show provider names (“OpenRouter”) instead of internal ids.

Before you upgrade. This version adds to your local database. If you later go back to rc.2 or earlier, that version starts with an empty database (the old file is kept as a backup).

1.0.0-rc.2 — 1 October 2026

App settings. Every saved app has a Settings page: choose which model answers its main and quick requests (or follow your active model), see its token use for the last week and set a daily limit for cloud models (models on your computer are never limited), browse, export or clear the data it keeps, revoke the sites and model providers it may use, and read a week of its activity. Prompts, replies and saved values are never recorded.

Starter apps in your language. The eight starter apps show their text in the interface language, with numbers, dates and currencies formatted to match.

Fixes. White-label branding works with the new design again: the Branding form starts from the current colours, a brand's colour reaches every accent, and turning branding on or off takes effect properly. Word documents keep characters like “&” and “é” when added to Documents.

Security. On Linux, pages can no longer open WebRTC connections: the webview starts with WebRTC off. It remains a known issue on macOS, which has no webview setting for it.

1.0.0-rc.1 — 30 September 2026

A new look. One design, dark or light (or following your system), replaces the theme gallery and theme files. The rail is labeled and carries the name and mark, and you can still pick the main colour for each mode. Ideas moved from the rail to the new-chat screen, under the message box, where picking one fills in the prompt.

Apps. Save a page the assistant built as an app and open it from the rail or from Your apps on the new-chat screen; it keeps working after you delete the chat. Eight starter apps come built in. Apps — and pages in a chat — can keep a little data between launches, take a few settings you fill in, and, only after you allow it, ask your AI model: the prompt names the provider and says when text would leave your device, and the page gets text in and text out — no tools, chat history, memory, or documents.

Workflows. Routines that fetch pages, search the web, have the model summarize, and save the result as a document — run now or on a schedule. Turning a schedule on shows everything the workflow may do on its own for you to approve, and a run that needs more pauses and asks.

Known issue: on macOS and Linux, a page's script can still open a WebRTC connection, which a page's content security policy doesn't cover. Windows blocks it in the webview itself; the macOS and Linux fixes haven't been built and verified yet.

0.1.0-rc.8 — 27 September 2026

Every page on the rail now shares one layout: a title, one line on what the page is for, the main action on the right, and a folded How this works for the longer explanation. Documents, Library, and Connectors are a list beside the selected item's full detail, Memory puts pending suggestions first, and Ideas uses the full width for its grid. The new layout is translated into all eight interface languages.

A page that needs data from many sites — a news reader, say — can now be allowed to reach any public site, for this session or always, instead of one site at a time. Every other rule still applies: https only, no local or private addresses, no cookies or credentials, the same limits, and nothing at all in local-only mode. And when an allowed site redirects to a different one, Conduit now asks about the new site instead of leaving the page with no data.

0.1.0-rc.7 — 26 September 2026

A new side rail holds Chats, Ideas, Documents, Library, Connectors, Memory, and Settings, and each opens as a full page instead of a pop-up — a reply keeps running while you're on another one. Settings is smaller as a result: providers, chat, web search, workspace, appearance, branding, privacy, and about. The panel beside the chat became an inspector with Page, Activity, and Sources tabs, so a finished reply shows one line — "2 steps · 1 site" — rather than a card per tool call, and the detail is one click away. The composer's row of icons became a single + menu, with whatever is switched on shown as chips, and a chat you've left shows "Running" or "Needs you" in the list.

Ideas is a page of twenty-one tested starting points — a pomodoro timer, a budget tracker, flashcards, Snake — that fill in a prompt you can edit before sending. An idea that needs something you haven't set up says so, and which ones you've tried is remembered on your device only.

HTML pages can fetch live data, from sites you allow. The page still has no network connection of its own: its requests go to Conduit, which asks you about each new site — once, always for this page, or not at all — and then makes the request itself, over https, without cookies or credentials. The globe button on a page lists every site it asked for and every request it made. Local-only mode refuses all of it, and one setting turns it off. Forms in HTML pages also work now; before, pressing a submit button did nothing.

Fixed: local-only mode now refuses cloud providers everywhere. Before, a chat that used tools — most chats, since the built-in tools are on by default — could still reach a cloud provider with local-only mode on. Plain chat, tool-using turns, image generation, and context compaction are all checked now.

0.1.0-rc.6 — 22 September 2026

Conduit can now chat with your own files. A new Documents collection in the sidebar takes plain text, Markdown, CSV, Word documents, and PDFs — drop them anywhere on the window, or attach a collection from the composer, and the assistant searches it while answering. A reply names the passages it drew on, and clicking one shows the exact text. Search combines meaning-based matching with exact keyword search, because each catches what the other misses. PDFs are parsed on your own machine; a scanned PDF with no text layer yields nothing, and the app says so rather than guessing. Indexing sends a document's full text to whichever provider embeds the collection — OpenAI, Gemini, or OpenRouter — unless that provider is Ollama, which stays local. Consent is asked per provider before the first document, remembered, and can be withdrawn from Documents at any time; with local-only mode on, a collection needs a local provider. Until you create a collection, chat is unchanged.

Ask for a picture and, on OpenAI, Gemini, or OpenRouter, Conduit generates one and saves it locally with the conversation rather than linking to a provider URL that can expire. Each image is billed by the provider, so Conduit asks once before the first one; asking about image generation doesn't trigger the prompt. Providers without an image endpoint are unchanged.

The prompts and resources a connected MCP server offers are now reachable from the composer, not just its tools: a prompt picker fills in the arguments a prompt declares and drops the result into the composer to edit before sending, and a resource picker attaches a document to the next message only, redacted and checked before the model sees it. A saved prompt with {{variables}} now asks you to fill them in when you insert it, instead of leaving the braces in the composer for you to find.

0.1.0-rc.5 — 15 September 2026

Six more providers shipped — xAI, Z.ai, Moonshot AI, Qwen, Together AI, and Fireworks AI — bringing the built-in total to 17. Anthropic and OpenAI gained a base URL field too, so either can point at a compatible endpoint such as a LiteLLM proxy, and hosted web search stays limited to the official endpoints.

Themes go beyond colour now: a theme sets type, corner radii, borders, elevation, and motion as well as the palette, and nine built-in themes are available from Settings → Appearance — Orange Charcoal (default), Orange-Dark, Terra, Amber Terminal, Green Phosphor, Amber Paper, Graphite, Editorial, and a High Contrast option built to AAA. A .theme.md file extends a built-in theme with your own colours — no CSS, no URLs. A separate Reading font setting picks the face for assistant replies.

Automatic updates are now optional. Settings → About offers only-when-you-check (the default, unchanged), notify-when-available, or install-on-quit; turning checks off still means no network request at all, and automatic install isn't offered for the Linux .deb. The sidebar can now be resized or reset with a double-click, a settings button in the title strip jumps to any section, typing in Settings search narrows the section list, and a keyboard shortcuts sheet (Ctrl+/, ⌘/ on macOS) lists everything. The artifact panel can be expanded (Ctrl+Shift+E), and while the assistant writes a document you can watch its line count and size grow, with a button to pick up where it stopped if a long one runs out of time mid-turn.

Fixed: a cloud provider chosen before the provider list finished loading could leave local-only mode on and the first message failing with an error about a setting nobody had seen; the model menu no longer mislabels priced cloud models as self-hosted; a narrow window's sidebar and artifact panel now open as overlays instead of becoming unreachable; and Ollama now receives tool definitions, so it can use tools at all.

0.1.0-rc.4 — 12 September 2026

First-run setup now opens by asking for your language, theme, palette, and text size, so the rest of it is readable before it explains anything, and ends on a review of what was actually configured. The interface is available in German, Spanish, French, Japanese, Korean, Brazilian Portuguese, and Simplified Chinese, alongside English, from Settings → Appearance, or by leaving it on System to follow the OS. Japanese, Chinese, and Korean text is drawn with the platform's own font for that language, and dates, times, number grouping, file sizes, and sorting follow the language you picked rather than the machine's region.

Fixed: choosing a cloud provider while local-only mode was on used to fail the first message with an error naming a setting nobody had seen — it now turns local-only off and says so. The first-run screen could render above the top of the window with no way to scroll to it; errors during setup were silent; the allowed- and blocked-domain placeholders showed a literal &#10; in German, Spanish, and French instead of a line break; truncated text now shows its full value on hover; and the connector consent dialog's description of a tool's permission level is built from the permission and the tool's own description now, instead of a fixed backend sentence, so it can be translated.

0.1.0-rc.3 — 6 September 2026

The agent release. A turn can now be steered while it runs: follow-up messages queue while a turn is in flight, a run can be interrupted mid-loop to redirect it, and tool approvals can be remembered for one conversation or permanently — except sensitive tools, which are never remembered and always ask. The model can also render a native form to ask a question instead of guessing.

Three larger pieces landed alongside it. Remote MCP over streamable HTTP, including OAuth 2.1 with dynamic client registration and tokens held in the OS keychain, plus read-only search of the official MCP registry with one-click install; SSE-only servers are refused with a reason rather than failing later. Agent Skills as SKILL.md packages enabled per conversation, with metadata read first and the instruction body only on enable — Conduit does not execute a skill's scripts/, they are listed as documentation. And persistent memory, where the model proposes a fact through a tool but it is stored as pending and never injected into a later prompt until you save it; everything stored is listed, editable, and deletable in Settings, and encrypted at rest only when encryption at rest is switched on, which is off by default.

A context gauge now measures the fill of the next request being built rather than summed turn usage, and compaction journals older turns into a summary as a conversation nears the model window — raw messages are never deleted; compaction applies when the request is assembled. Conversations gained pinning, archiving, and one level of folders, and a segment timeline renders thinking, tool calls, and answers in the order they actually happened. Always-show-reasoning is on by default.

0.1.0-rc.2 — 2 September 2026

Web search became pluggable: a local backend with a DuckDuckGo default plus Tavily, Brave, and SearXNG, alongside Anthropic hosted search, selected under a dedicated Auto / Hosted / Local setting. Image attachments are now sent to models that accept them, and conversations can be exported as Markdown or JSON.

Chat markdown gained Mermaid diagrams and KaTeX maths. Messages can be edited and resent, including forking mid-thread, and each conversation can carry its own generation controls and custom instructions. Workspace folder tools are opt-in, with a chip in the chat showing when a folder is bound. This is also the release where white-label branding arrived in both forms — runtime branding from Settings, and build-time branding for a build made from source.

0.1.0-rc.1 — 22 August 2026

The first packaged release candidate: unsigned installers for Windows, macOS on both Apple silicon and Intel, and Linux as .deb and AppImage, with update payloads signed by an Ed25519 key. It carried the initial public licensing under AGPL-3.0, contributor documentation, and third-party attribution.

Download the current release

New here? Start with the documentation, see what stays on your machine, or read how to demo an MCP server to a customer.