Lorca for Android 1.0.4

• A chat that asks for a channel gets one reply, "Pending approval", and once you allow it, the bot
  answers what people asked while it waited. A chat that never asks can be added by its id with
  Add Chat… on the channel's screen.
• Conversations a Telegram or Slack channel keeps show the service's mark on their avatar, so they
  read apart from the bot's own DM.

Lorca for Android 1.0.3

• A channel answers only in the chats you allow. Anyone can message a Telegram bot or add it to
  their group, so a chat you haven't allowed now asks first: the channel's row says A chat is
  waiting, and its screen lists it with Allow and Ignore. Until you allow it, the bot does nothing
  there. A channel that listened in every chat its bot was in now listens nowhere until you allow
  its chats, and each chat asks once someone writes there.
• Tapping an item in the Attention list that waits on a review opens its chat with the review on
  top, so you can approve or reject it right away.

Lorca for Android 1.0.2

• A Telegram channel that listens for mentions or hashtags tells you when it can't hear them. While
  the Telegram bot's privacy mode is on, Telegram sends it only replies and commands in groups. The
  channel's row in a DM's Details now says Privacy mode on, and its screen says what it misses and
  how to turn privacy mode off.
• On Android, a message you send rests below the chat bar while the reply comes in, instead of
  scrolling up under the status bar.
• On Android, Install Update shows the download's progress in a dialog until the app restarts, and
  says so when the download or install fails.

Lorca for Android 1.0.1

• Update Lorca on every Device: the phone app, the Mac app, the Windows and Linux app, and each
  computer that runs the command line. Devices now sign what they send each other through the relay,
  so a relay can no longer send work to your Runners in another Device's name, and unpairing a
  Device gives your account a new key, so the Device you unpaired cannot read what your account
  writes after that. An older Lorca cannot read what an updated Device sends, and once your Devices
  have updated, Lorca's relay asks older versions to update before they sync.
• A bot's Access changes from the phone: the Access row in a DM's Details, or Edit Access… on a
  bot's request for more access. Each change saves at once.
• A bot writes an email or a Slack message as a draft card in the chat, editable in place, and
  nothing goes out until you tap Send. What a routine leaves for your review waits in the chat the
  same way.
• What bots publish shows as a card in the chat, and Details lists a chat's outputs and tasks. New
  Task in Details adds a task.
• A secret a bot asks for goes on a card you fill in from the phone, sealed to the bot's Runner; it
  never enters the chat. The Runner's screen in Settings lists its secrets, without their values.
• Tapping a routine opens its own screen: how it stands and how to fix it, Run Now, Pause or Resume,
  Limits, its schedule and timezone, and missed runs. A routine that needs you says so in Details.
• Sidebar sections, and Hide and Mute in a chat's menu. A muted chat sends no push, and neither does
  the chat you have in front on a computer.
• Paste an image into the message box.
• Give Feedback in a DM's Details, project context in a group's Details, a bot's channels, browser
  profiles, and coding agents' cards with their transcripts, as on a computer.
• Share a bot as a template link, and open one to make a bot from it.
• Any Device renames any Device: Rename… on its screen in Settings.

Lorca for Windows, Linux, and macOS 0.1.9

• A channel answers only in the chats you allow. Anyone can message a Telegram bot or add it to
  their group, so a chat you haven't allowed now asks first: the channel's row says A chat is
  waiting, and its sheet lists it under Asking to Chat with Allow and Ignore. Until you allow it,
  the bot does nothing there. A channel that listened in every chat its bot was in now listens
  nowhere until you allow its chats, and each chat asks once someone writes there.
• Clicking an item in the Attention list that waits on a review opens its chat with the review on
  top, so you can approve or reject it right away.

Lorca for Windows and Linux 0.1.8

• A Telegram channel that listens for mentions or hashtags tells you when it can't hear them. While
  the Telegram bot's privacy mode is on, Telegram sends it only replies and commands in groups. The
  channel's row in the bot's Details now says Privacy mode on, its sheet says what it misses and how
  to turn privacy mode off, and the bot tells you when it sets up or changes such a channel.

Lorca for Windows and Linux 0.1.7

• Update Lorca on every Device: the Windows and Linux app, the Mac app, the phone app, and each
  computer that runs the command line. Devices now sign what they send each other through the relay,
  so a relay can no longer send work to your Runners in another Device's name, and unpairing a
  Device gives your account a new key, so the Device you unpaired cannot read what your account
  writes after that. An older Lorca cannot read what an updated Device sends, and once your Devices
  have updated, Lorca's relay asks older versions to update before they sync.
• Each bot has Access, which only you change: how far it may use each plugin (Read only, Read and
  draft, Read and write, or No access), which of its tools it may call, and whether it may read or
  change files and run commands on its Runner. When a bot reaches a limit, the chat offers Edit
  Access… to change it. A bot you never set keeps full access.
• A bot writes an email or a Slack message as a draft card in the chat, with its recipients,
  subject, text, and attachments editable in place, and nothing goes out until you press Send.
  Access › Messages turns drafts off for a bot, as Always Send does on a Slack card.
• Routines no longer refuse what Auto-review would ask you about: they leave it for your review. The
  chat says so, the inspector lists what waits, and Approve runs exactly what the sheet shows, edits
  included.
• Bots publish what they make, such as a report, a screenshot, or a chart, with what they checked:
  Passed, Failed, or Not verified, and the command they ran. An output shows as a card of its own in
  the chat, and the inspector lists a chat's outputs.
• Tasks: work that outlasts one turn, with an owning bot, what done looks like, the next step, and
  the evidence once it's done. A chat with tasks has a Tasks section in the inspector, and File ›
  New Task… adds one.
• A bot that hands work to a teammate gets the result back in its chat when the teammate finishes,
  gets stuck, or is stopped, also from a Runner that was offline when it asked.
• One list of what needs you: the coordinating bot keeps the reviews, blockers, and promises your
  bots report, with a short brief, behind a toolbar button that appears while something waits.
• A group chat keeps a shared project context beside its transcript: a brief, goals, constraints,
  decisions, facts, links, and files, which every bot in the group reads. It's the Project section
  of the group's inspector.
• Skills: Save as Skill… on a bot's reply, or Save as Standing Instruction… on your own message,
  writes a reusable instruction from the messages you pick, and it is used once you save it. Give
  Feedback… on a bot's message teaches the bot, which suggests changes to its routines and skills
  for you to accept.
• Workflows set up an outcome in one place, such as meeting preparation, inbox triage, or watching a
  repository: the bot, the accounts it needs, and a sample run before it runs on a schedule. They're
  first in the marketplace, and on onboarding's last page.
• Share a bot as a link: File › Share as Template… shows exactly what the template holds, keys and
  passwords removed, and the link opens a page that shows the bot and opens it in Lorca. Neither the
  relay nor lorca.app can read it.
• Limits on a bot's turns, a routine's runs, and a task: spending, tokens, run time, retries, and
  plugin calls, with Resume when work stopped at one. Each plugin account has a call limit all bots
  share.
• Browser profiles keep a bot's sign-ins on its Runner. Take Over gives you the bot's browser and
  Return to Bot hands it back, and Record teaches the bot a browser workflow by doing it once.
• Gmail, Google Calendar, Google Drive, and Slack are in the marketplace, each with as many named
  accounts as you want, such as a Work Gmail beside a Personal one.
• A bot asks for a password, a key, or a code on a card, and the value goes sealed to its Runner. It
  never enters the chat or the model. Settings › Secrets lists them, without their values.
• Routines keep their timezone, run once when their Runner comes back from missed times (or skip
  them), and say when they need you. A bot can also set a one-time reminder, watch a pull request,
  run around your calendar's events, or run on a webhook, hosted on Lorca's relay.
• A bot can listen on a Telegram or Slack account and answer people in their thread. The Feedback
  collector workflow files community feedback as GitHub issues.
• A bot can run Claude Code or Codex on its Runner and supervise it, and the chat shows the run as a
  card with its transcript and its questions.
• Sidebar sections, and Hide and Mute in a chat's menu. A muted chat sends no notification and stays
  out of the taskbar badge, and the phone gets no push for the chat you have in front on a computer.
• Stop also stops the work a turn handed off, and ends a turn within about 15 seconds whatever its
  tools do.
• Any Device renames any Device: Rename… in its Settings › Devices pane.
• A message you send while a bot waits on a background command, or on a browser you took over,
  reaches the bot at once.
• On a Windows Runner, a bot's commands run in a terminal, so they can go to the background, take an
  answer, and show in Running tasks, and Stop ends one that held the turn.
• Claude Haiku 5.5, and Sonnet 5.5's lower cache read price.
• Only an app on this computer that holds Lorca's token can connect to its command line.
• Typing with an input method, such as pinyin, keeps an @mention in the message box styled.
• Settings › Providers has Review Models: pick the model Auto-review runs on each connected
  provider instead of its small, fast default, including a decision model such as OpenCode Zen's
  Jev 1.13, which only allows an action or names what it could harm. Settings › Auto-review has
  Reviews with, to review with another connected provider than the bot's.
• Add Provider… offers decision APIs for Auto-review: OpenRouter Decisions, OpenAI Decisions, and
  TypeSafe, and a custom provider can speak System One or OpenAI Decisions. Bots never run on
  them.

Lorca for Windows, Linux, and macOS 0.1.10

• A chat that asks for a channel gets one reply, "Pending approval", and once you allow it, the bot
  answers what people asked while it waited. A chat that never asks, such as a group where the bot
  only hears replies to it, can be added by its id with Add… on the channel's Chats row.
• Conversations a Telegram or Slack channel keeps show the service's mark on their avatar, so they
  read apart from the bot's own DM.

Lorca CLI 0.1.15

Install:

curl -fsSL https://lorca.app/install-cli.sh | sh    # macOS and Linux
irm https://lorca.app/install-cli.ps1 | iex         # Windows, in PowerShell

A CLI installed this way updates itself; lorca update installs this release now.

Lorca CLI 0.1.14

Install:

curl -fsSL https://lorca.app/install-cli.sh | sh    # macOS and Linux
irm https://lorca.app/install-cli.ps1 | iex         # Windows, in PowerShell

A CLI installed this way updates itself; lorca update installs this release now.