BoltAI Updates & Release Notes

This release adds a faster model selection workflow, more capable MCP connector authentication, and a long list of reliability fixes across chat, menus, windows, and the chat renderer.

New

• Added support for searching the Model Picker across all AI providers.
• Added support for favoriting and starring models so they stay near the top of the selection list.
• Added support for MCP connectors that authenticate with custom headers.
• Added configurable maximum tool-call limits.
• Added chat and message date visibility in the conversation UI.
• Added support for merging a Project System Prompt with an Agent or Chat System Prompt when both are set.
• Added chat duplication.
• Added GPT-5.5 support for Codex.

Improved

• MCP OAuth sign-in and reconnect flows are more reliable, with better recovery for remote connectors and local confidential-client setups.
• The Model Picker works better with large provider catalogs, with faster search, clearer recovery, and better Settings fallbacks.
• Streaming reasoning output is easier to follow while a response is still generating.
• Provider model management and hosted model catalog refresh are more reliable.
• The chat title area is more stable on macOS 15.
• The instant chat bar has better placement and more reliable deep-link behavior.
• BoltAI does a better job remembering the size and position of the instant chat bar and main app window.
• MFA enrollment now renders QR codes natively for a more reliable setup flow.
• Time-sensitive web search prompts now use clearer UTC date context.

Fixed

• Fixed crashes that could happen while closing the chat view or menu bar UI.
• Fixed crashes that could happen while the chat input view was being torn down.
• Fixed a crash when global shortcuts open chat windows.
• Fixed UI freezes and race conditions when creating new chats on older macOS versions.
• Fixed a crash when the permission overlay hides after returning from System Settings.
• Fixed crashes that could happen when using the Workflow Palette repeatedly.
• Fixed crashes that could happen when opening Settings, Multi-Agent Chat, or instant chat from menus.
• Fixed crashes that could happen when copying selected text from the chat view.
• Fixed the inspector tab layout crash.
• Fixed custom provider model fetch crashes.
• Fixed instant chat draft reset behavior.
• Fixed instant chat deep links.
• Fixed duplicate main window cleanup on macOS 15.
• Fixed AI service key cache race conditions.
• Fixed a crash when viewing Recently Deleted items from Project view.
• Fixed Instant Chat not resetting correctly when it was set to reset immediately.
• Fixed compose thinking popover crashes when using custom configuration actions.

New

New in 2.10.0

• New: Added new providers including Hermes Agent, OpenClaw, OpenCode, MiniMax, Xiaomi MiMo, and Z.AI.
• New: Added MiniMax media tools in chat, including image generation, image edit, music generation, and lyrics generation.
• New: Added MiniMax text-to-speech support with voice picker.
• New: Added xAI text-to-speech support with streaming playback.
• New: Markdown soft line breaks now render correctly in chat.

Improved

• Improved: Local and OpenAI-compatible provider setup is more capable, including LM Studio API key support and better OpenCode password auth handling.
• Improved: Generated media attachments are more reliable, with better saving and rendering for images and audio in chat.
• Improved: Strengthened CryptoV2 key handling and clarified provider error messaging.
• Improved: Hermes Agent responses can now include tool URL citations in chat.
• Improved: The update menu now gives clearer status for available and downloaded updates, with quicker install actions.
• Improved: OpenCode chat binding and provider-default handling are more stable, especially around new chats and clearing defaults.
• Improved: Improved sidebar and chat switching performance.
• Improved: Sync is more resilient to referential integrity and dependency repair issues.

Fixed

• Fixed: Fixed attachment photo ingestion and compose attachment preparation races.
• Fixed: Fixed XLSX extraction for workbook metadata relationships.
• Fixed: Fixed a bug when using Anthropic or Claude Code providers with native skills.
• Fixed: Fixed duplicate main window reopen behavior on macOS 15.

build 63

• Fixed: Compose undo and redo now behave much more naturally, with native AppKit-style undo grouping in the prompt editor.
• Fixed: Drafts, cursor position, and undo history are now isolated per chat, so switching chats no longer mixes compose state together.
• Fixed: Prompt-library inserts and other programmatic text inserts now undo and redo like normal editor edits.
• Fixed: Attachment sending is more reliable in compose, including locked-send retry flows and chat switching with unsent attachments.
• Improved: Workflow, system prompt, and title prompt editors now save more safely, keep the sheet open on failure, and show clearer inline errors.
• Fixed: Local instant dictation now uses the correct OpenAI-compatible /v1 endpoints for Ollama and LM Studio.
• Improved: Reply-finished notifications now include a sound again.

build 62

This is one of the biggest BoltAI updates yet, with a major wave of improvements across security, sync, provider setup, model handling, voice workflows, and overall app polish.

• New: Added multi-factor authentication support on macOS.
• New: Added Anthropic prompt caching controls.
• New: Added native instant dictation cleanup.
• New: Added Mermaid preview support in the renderer.
• Improved: Authentication is more reliable, with stronger sign-in, reauth, and account recovery flows.
• Improved: Cloud Sync is more reliable around expired sessions, reauthentication, and MFA-required states.
• Improved: Amazon Bedrock setup is more capable, with better service configuration, credential resolution, and profile handling.
• Improved: Azure OpenAI setup is smoother, with custom base URLs, deployment URL import, lighter validation, and better API version recovery.
• Improved: Codex and Claude Code authentication recovery is more reliable.
• Improved: Claude Code now supports Claude Sonnet 4.6.
• Improved: AI service management is more capable, with bulk editing, better model catalog persistence, and improved preservation of enabled model selections.
• Improved: Voice workflows are more polished, with better recent dictation handling.
• Improved: Security is stronger, with libsodium-based crypto defaults and more robust licensing and JWT validation.
• Fixed: Fixed Anthropic prompt caching metadata propagation and usage reporting.
• Fixed: Fixed XLSX extraction and handling of unsupported legacy Office files during import.
• Fixed: Fixed sidebar chat rename issues.
• Fixed: Fixed prompt title generation behavior.
• Fixed: Fixed live chat web fetch source persistence.
• Fixed: Fixed catalog sync identity collision issues.
• Fixed: Fixed auth-related race conditions during initial sync and reauthentication.
• Improved: Cloud Sync pause and recovery messaging is clearer, with better guidance for retrying sync, exporting diagnostics, and understanding when local data is safe.
• Improved: Local diagnostics exports now skip Setapp auth diagnostics in standalone builds.
• Improved: Sidebar folder hierarchy spacing is more consistent and easier to scan.
• Improved: Added a new chat toolbar title rendering mode flag for continued title bar refinement on macOS.
• Improved: Developer settings now provide clearer local data wipe warnings and safer recovery guidance when Cloud Sync is unavailable.
• Improved: BoltAI now uses the published Anthropic fork package with improved stream handling coverage.
• Fixed: Fixed Cloud Sync ordering for title generation provider settings to avoid foreign key sync failures.
• Fixed: Deleting an AI service now clears related custom title generation settings before sync.
• Fixed: Foreign key sync validation failures are now reported to Sentry with safer diagnostic metadata so we can investigate issues faster.

Build 60

• Added: Message diagnostics now show response timing and generation speed, with support for always showing tokens-per-second when enabled in settings.
• Improved: Chat images are easier to work with. Saving, dragging, and download flows are more reliable, including images and files generated from in-chat data URLs.
• Improved: Code blocks and SVG outputs now render more cleanly, with better save actions for generated artifacts.
• Improved: Stored API keys now use stronger encryption settings by default, and Security settings can prompt you to upgrade older encrypted services.
• Fixed: Cloud Sync is more reliable around session refresh, first-sync recovery, and chat paging, reducing cases where sync retries could fail or leave partial local state behind.
• Fixed: Improved OpenRouter web plugin availability so compatible tools are offered more consistently in chat.
• Fixed: Inline math parsing is now safer by default and correctly follows your rendering setting, reducing accidental math formatting in normal text.
• Improved: License refresh now provides clearer feedback in Settings and the upgrade flow.

Build 59

Improved

• Added support for a new title generation mode "prompt": automatically generate a title using user prompt - no LLM involved.
• Better Cloud Sync reliability for attachments in some edge cases.

Fixed

• Improved reliability for Claude and Anthropic chats that use Search, so older conversations are less likely to fail when you continue the thread later.
• Resolved an issue where some Claude and Anthropic chats with tool use could get stuck with tool-history errors instead of continuing normally.

Build 58

• Added: Support for GPT-5.4 in Codex CLI.
• Added: Support for Claude Opus 4.6 in Claude Code.
• Improved: Better Claude Code integration in provider setup and model selection.
• Improved: Cloud Sync now pauses more safely when your app version is too old or local data integrity issues are detected, helping prevent bad sync states.
• Fixed: Better handling for speech-to-text model IDs in OpenAI-compatible flows.
• Fixed: Attempted a fix for an issue reported by some users on older macOS versions where creating a new chat could produce duplicate New Chat entries.

Build 55

New

• New: GPT-5.3 Codex Spark support.
• Added GPT-5.3 Codex Spark and improved default model prioritization.
• New: Search plugin parity.
• Added first-class Perplexity and You.com search tool support.
• New: Sidebar toggle shortcut.
• Added a quick sidebar toggle command with Cmd+Option+S.
• New: Starter pack install flow.
• Improved onboarding with a clearer starter pack setup path.

Improved

• Improved: Parallel chat streaming stability.
• Isolated streaming state per chat session to prevent cross-session interference.
• Improved: Anthropic/Bedrock behavior.
• Better Claude thinking handling and cleaner Anthropic control behavior.
• Improved: Workflow palette behavior.
• Active-screen behavior now works correctly while preserving per-screen restore.

Fixed

• Fixed: Voice recording/transcription UX.
• Polished compose recording overlay and transcription error presentation.
• Fixed: Low-res App Icon.
• Fixed the issue with the low-res default app icon.
• Fixed: Grok/xAI request reliability.
• Migrated xAI/Grok flows to the Responses tool path to avoid deprecated endpoint failures.
• New: xAI native tools.
• Added xAI Web Search, X Search, and Code Execution tool wiring in chat transport and tool builders.
• Improved: Tool execution rendering.
• Provider-executed tool calls now render more consistently, including xAI code-execution outputs and fallback detail views.
• Improved: Provider error diagnostics.
• Error messages now include richer raw-response context and clearer guidance for deprecated xAI Live Search behavior.
• Fixed: Settings sync edge cases.
• Hardened account/plugin settings sync with stricter canonical fetch checks, safer merge-upsert fallback paths, and duplicate-row intent preservation.
• Fixed: Sidebar multi-select move action.
• Context-menu move operations now correctly apply to the current multi-selection and disable invalid destinations.

Build 54

• New: Appearance controls. Added dedicated appearance settings so you can customize app icon style, sidebar presentation, and chat content width for a more personalized layout.

• New: Chats inside folders in the sidebar. You can now browse and navigate chats directly inside folders from the sidebar for faster project-based navigation.

• New: GPT-5.3 Codex support (Codex CLI Provider). Added support for the latest GPT-5.3 Codex model.

• New: Title generation controls. Added title generation settings and improved per-chat custom configuration behavior. Go to Settings > General to customize.

• New: Expanded import/export. Improved import and export reliability, with better support for ChatGPT imports and expanded coverage for Claude, Grok, Mistral, and BoltAI v2 data.

• Cloud sync improvements. Sync is now more reliable for large histories and after signing in from Settings.

• Cloud sync status clarity. Improved sync progress and error messaging so it is easier to understand what is happening.

• Navigation and responsiveness improvements. Smoother sidebar behavior, better chat-in-folder navigation, and reduced settings-related UI churn.

• Fixed: Edit/regenerate reload issue. Resolved a case where edit-regenerate could trigger unnecessary chat reloads.

• Fixed: Sync retry edge cases. Fixed several sync scenarios that could fail, stall, or require repeated retries.

• Fixed: Sync consistency edge cases. Resolved issues that could occasionally leave cloud sync in an inconsistent state.

New

• NEW: Moonshot AI provider support. Added Moonshot AI as a selectable provider.
• Codex tools, fixed. ChatGPT OAuth (Codex) now supports custom function tools (including MCP) with correct tool-call streaming and input/output handling.
• Codex tool-call safety. Prevents mismatched tool-call state when switching providers by dropping unsupported tool parts before requests.
• Remote MCP reconnects on launch. OAuth MCP servers now reconnect reliably at startup, with proper auth-mode persistence for custom connectors.
• OAuth token lookup improvements. Better matching and handling for MCP OAuth tokens to reduce unnecessary re-auth prompts.
• Enter-to-send preference. Added a composer option to send on Enter.
• Code fence inline detection fix. Improved inline detection for fenced code blocks.

New

Improved

• Gemini CLI web search. Added provider-native Google Search support with search indicators and source citations, so Gemini CLI results display grounded links in the UI.
• Codex web search compatibility. Added tool wiring and status indicators so Codex CLI web search shows up cleanly in chat output.
• Provider-aware tool UI. Tool selection now adapts to Gemini CLI (search-only) and clarifies Copilot tool availability.
• Improved math rendering. Better inline LaTeX handling for typical model output using single-dollar delimiters.

New

• New: Google Vertex AI. Configure service account JSON, project, and location to use Vertex models.
• New: Chat export/import. Export a chat to JSON or import it from the app menu.

Improved

• Improved: Tool history reliability. Anthropic server tools are sanitized to avoid malformed replay.
• Improved: Usage reporting. OpenAI‑compatible providers now surface usage metrics consistently.

Fixed

• Fixed: Gemini image follow‑ups. Thought signatures are preserved and images are normalized for multi‑turn use.
• Fixed: Native tool availability. Provider-native tools now show up correctly (including new Vertex tools), with tool selection remaining per-chat.