Lori v1.5.5

Native Claude and Codex conversations, complete source-control workflows, and safer parallel agent sessions in a calmer Lori workspace.

Added

• Structured Claude and Codex conversations — work with both agents in a native conversation interface with live responses and reasoning, tool activity, plans, approvals, forms, context usage, subagents, and conversation controls.
• Agent coordination — Claude and Codex tabs in the same Lori session can discover peers, exchange messages, delegate work to new tabs, and return results. Settings includes spawn approvals, active-tab limits, and pending requests.
• GitHub, GitLab, and Bitbucket workflows — connect repositories, clone projects, create and review pull or merge requests, follow request status, and merge completed work from Lori.
• Flexible worktree sessions — create a new branch, attach an existing branch, check out a pull or merge request, or import an existing linked worktree. Custom worktree locations and sparse checkout are also supported.
• Lori account sessions — sign in through Lori Dashboard, inspect device and session details, manage connected devices, refresh verification, and continue through temporary offline periods.
• Managed Claude runtime — Lori can securely install, verify, repair, and update the runtime used by structured Claude conversations.

Improved

• New sessions — the redesigned prompt-first composer lets you choose an agent, enter the first request, add an optional title, and configure worktree options before starting.
• Conversation streaming — responses now arrive smoothly without per-word flicker, while incomplete Markdown, links, lists, and code blocks remain stable and readable as they stream.
• Reasoning display — adjacent thoughts are grouped into a calm live view, completed reasoning collapses into an accessible summary, and light-theme shimmer keeps the underlying text readable.
• Approvals — simultaneous permission requests are presented one at a time in arrival order, with a clear count of requests still waiting.
• Source control — change-request status loads without holding up the Changes panel, request links stay matched to the active branch, and hosted Git operations are more predictable across forks and remotes.
• Worktree management — session creation, import, fan-out, rename, base updates, deletion, and interrupted-operation recovery now use durable lifecycle states with clearer repair guidance.
• Navigation and files — the sidebar, titlebar, tabs, panels, status area, and composer share a cleaner desktop layout with smoother reduced-motion-safe resizing and richer file-type icons.
• Organization policies — managed providers, plugins, and skills stay visibly attributed and continue following the last verified policy during temporary server outages.
• Usage insights — dashboard telemetry views and Enterprise usage synchronization provide clearer activity and usage reporting.

Fixed

• First prompt delivery — the opening request is delivered exactly once and restored as an editable draft when Lori cannot confirm delivery, preventing lost or duplicated first turns after remounts, reconnects, or restarts.
• Agent runtime recovery — structured conversations recover more reliably from crashes, restarts, stale saves, queued prompts, authentication changes, and switches between terminal and structured interfaces.
• Worktree ownership safety — imported worktrees remain user-owned and are never force-removed during rollback or session deletion; local branches are preserved unless deletion is explicitly requested.
• Worktree consistency — legacy databases, hidden workspace collisions, checked-out branch eligibility, stale Git state, and interrupted creation or deletion no longer leave sessions pointing at unsafe or missing paths.
• Pull and merge request accuracy — Lori no longer selects an unrelated request with the same branch name, silently relies on failed remote fetches, or pushes a checked-out request to the wrong source branch.
• Workspace removal — removing the active workspace returns to the home screen instead of leaving an empty application shell, while its sessions, histories, branches, and worktrees remain available when re-added.
• Agent coordination delivery — peer messages are durably queued, delivered once, and never replayed when delivery is uncertain. Typed terminal input is protected, stale delivery claims expire safely, and completed messages are not later marked failed.
• Peer-message trust — transcript text cannot impersonate another Lori agent; peer presentation now requires verified sender, recipient, and message identity.
• Enterprise offline enforcement — temporary connectivity failures no longer relax organization restrictions; the last verified policy remains active until its authority expires.
• Conversation rendering — blank lifecycle rows no longer hide active reasoning, streamed content no longer produces glitch-like animation, and completed-response actions remain hidden until hovered or focused.

Lori v1.5.4

Snappier Skills Explore installs and a simpler Settings header.

Improved

• Skills Explore — charts and search stay on screen while they refresh, so Add stays clickable; installed skills update right away, and reopening Explore doesn’t flash an empty loading state.
• Settings — the sidebar header is a single back control instead of a stacked title row.

Lori v1.5.3

Richer conversation handoffs when you switch agents — tool use, reasoning, and attachments come along, not just the text.

Improved

• Switch agents with full context — moving a session between Claude, Codex, OpenCode, Cursor, Antigravity, Pi, and the other supported CLIs keeps the conversation looking like the real chat, including tools and attachments, instead of a flat text paste.
• Simpler switch dialog — Sync always carries history forward; the extra “fresh session” choice is gone so it’s clearer what will happen.

Fixed

• Hand-offs into Pi, Codex, Cursor, OpenCode, and Antigravity open and show history more reliably, including chats that previously looked empty or missing messages after sync.

Lori v1.5.2

Desktop controls in Settings, better workspace icons, and a cleaner Mac titlebar.

Added

• Desktop settings — control close behavior, menu bar tray, notifications, launch at login, and idle session suspend from Settings.

Improved

• Workspace icons — Lori finds more project logos for nested frontend apps and monorepos, and sidebar avatars fill their tiles without an extra frame.
• macOS titlebar — traffic lights sit centered in the overlay header.
• Destructive confirms — delete and similar actions use a clearer danger button on dark UI.

Lori v1.5.1

Smoother Files and Changes workflows, plus a clearer path when no IDE is installed.

Improved

• Files panel — the in-app editor syncs more reliably with disk, fills the panel cleanly, and keeps theme-aware syntax highlighting readable in light and dark modes.
• Changes sidebar — changed files show as a simple flat list (like VS Code), so it’s easier to scan and act on what’s dirty.
• No IDE installed — a quieter chip in the UI opens a short help panel instead of feeling like an error.

Fixed

• Opening the IDE help panel no longer locks up the UI in a render loop.

Lori v1.5.0

Lori is now available on Windows, with the same in-app updates you already get on Mac and Linux.

Added

• Windows — download Lori for Windows from GitHub Releases, install it, and keep working with your agents in one workspace. After install, updates show up in the titlebar like on other platforms. Windows may warn on first open until broader Windows signing lands.

Changed

• Releases for every platform — each Lori release now ships Mac, Linux, and Windows builds together, so you can pick the installer for your machine from the same release page.

Fixed

• Windows terminal — agent output renders as a live terminal again instead of looking like frozen plain text, with clearer fonts and spacing in the packaged Windows build.
• Windows browser tabs — opening a browser tab no longer fails on Windows.
• Closing Lori on Windows — quitting from the window close button asks for confirmation so you don’t lose running sessions by accident.
• Agent tabs — closed tabs no longer come back as empty drafts after a refresh, and the “no CLI found” empty state no longer flashes while agents are still being discovered.

Lori 1.4.14

More reliable Connections provider sync and a sturdier Anthropic translation bridge.

Fixed

• Provider sync races — Connections config mutations are serialized so concurrent register/update/import flows no longer clobber each other; frontend register/sync edge cases from the providers audit are closed.
• Anthropic bridge errors — the translation bridge passes through HTTP status correctly, hardens SSE framing, and surfaces upstream failures more clearly instead of hanging or dropping the stream.
• Provider URL detection — clearer auto-detect for custom OpenAI-compatible endpoints when adding a provider.

Full changelog: https://github.com/loridotsh/lori/blob/main/CHANGELOG.md

Lori 1.4.13

In-app file editing, git-aware Files explorer, smoother terminal scrolling, and local HTML previews in the embedded browser.

Added

• Edit files in-app — the file viewer is now an editor: click the pencil (or open any text file) and type. Autosaves after a pause, ⌘S saves immediately, and Tab indents using the file's own style. Saves are crash-atomic (temp file + rename), keep the file's permissions and CRLF/LF line endings, and refuse to clobber: if the file changed on disk since you opened it, you choose to reload or overwrite. Editing stays disabled for binary, image, oversized (2 MB+), and non-UTF-8 files so a save can never destroy data.
• Git status in the Files explorer — changed files show status letters (M/A/D/R/U) and colored names, and folders with changes are tinted so you can spot dirty areas at a glance. New toolbar with collapse-all and refresh.
• Local HTML in the browser — open workspace HTML via file://, absolute paths, or ./ / ../ in the omnibox. Loads stay scoped to the session workspace so relative CSS/JS/images work without exposing the rest of the disk.

Fixed

• Terminal scrolling — smoother stick-to-bottom while agents stream output, native macOS trackpad scrollback, and fewer layout-driven jumps when resizing the sidebar or Changes panel. Hidden and inactive panes keep flushing PTY output so backpressure acks never stall.

Changed

• Confirm dialogs — cleaner, divider-free modal shell with neutral ink confirm actions (no saturated blue) for close-tab and other confirmations.
• Untracked file actions are unambiguous — untracked rows now show a "U" letter (previously the same green dot as added files), and deleting untracked files says "Delete permanently" everywhere, so it can't be mistaken for a reversible discard.

Full changelog: https://github.com/loridotsh/lori/blob/main/CHANGELOG.md

Lori 1.4.12

Pi coding agent support, one-click CLI install, and an update-available popup with release notes.

Added

• Pi — run the Pi coding agent in Lori alongside your other tools. Start and resume sessions, sync marketplace skills into ~/.pi/agent/skills/, see running/idle status via the Lori notify extension, and sync Connections providers into ~/.pi/agent/auth.json and models.json.
• Pi MCP marketplace sync — Lori writes enabled MCP plugins to ~/.pi/agent/mcp.json and ensures the community pi-mcp-adapter package is listed in Pi settings (installing it when the pi CLI is available).
• One-click CLI install — Settings → Coding tools shows an Install button for missing agents (official installer, npm, or Homebrew depending on the tool and OS), with a live log and copyable command.
• Update available popup — when a new Lori version is ready, a modal shows the release notes with Update & restart or Later (reminds again after a day).

Full changelog: https://github.com/loridotsh/lori/blob/main/CHANGELOG.md

Lori 1.4.11

Richer prompt composer slash commands and real media previews in the Files panel.

Added

• Slash command palette — type / in the prompt composer to insert agent CLI commands (Claude, Codex, OpenCode, Antigravity, Hermes, Cursor), ranked like @ file mentions.
• Skills in / menu — enabled marketplace skills for the active agent show up alongside built-in commands, so you can insert /skill-name without leaving the composer.
• Media previews — the Files panel previews images, video, audio, and PDFs instead of showing binary placeholders.

Changed

• Slash catalogs — per-agent command lists expanded to match current CLI docs, with fuller coverage for every supported agent.

Full changelog: https://github.com/loridotsh/lori/blob/main/CHANGELOG.md