SuperClaw Release v1.3.0

Highlights

• Faster Startup and Smaller Footprint
  Major runtime and service architecture optimizations deliver faster startup, lower resource usage, and a significantly smaller installation footprint.

• Enhanced File Protection
  Expanded privacy and file protection capabilities, including multi-file protection, sensitive folder management, masked-file previews, and improved protected-file workflows.

• More Powerful Local AI
  Added support for Qwen3.8-27B on a single Intel® Arc™ Pro B70, improved model routing and reliability, and support for Copilot responses-only models.

• Speech-to-Text (STT) and Text-to-Speech (TTS) I/O Support
  Added support for native voice interactions through new Speech-to-Text (STT) and Text-to-Speech (TTS) interfaces. Users can speak naturally to agents and receive spoken responses, enabling a more accessible and hands-free AI experience across supported workflows.

• Improved Installation & Upgrade Experience
  New in-app upgrades, hardware compatibility checks, launch options, flexible uninstall controls, and enhanced consent management.

• Smarter Agent Automation
  Schedulers now support editing, pausing, and resuming scheduled agents, with improved scheduling workflows.

• Security & Privacy Hardening
  Comprehensive security, privacy, and dependency updates, including penetration-test remediation, secure download validation, prompt-injection protections, and stronger protected-data safeguards.

Added

Scheduler Enhancements

• Edit, pause, and resume scheduled agents.
• New scheduler presets and improved scheduling workflows.

Enhanced File Protection

• Sensitive folder file-list management.
• Multi-file batch protection.
• Preview masked files directly in Chat.
• Improved PII restoration and protected-file workflows.

Installer Improvements

• In-app upgrade support.
• CPU and GPU compatibility checks.
• Launch after install and optional launch on reboot.
• Telemetry and crash-reporting consent experience.
• Independent uninstall options for app data and downloaded models.

Model & AI Capabilities

• Support for local Qwen3.8-27B deployment on a single Intel Arc Pro B70.
• Support for Copilot responses-only models.
• Context-budget awareness for agent runtime requests.
• Direct URL model downloads for selected packages.

GitHub & Sandbox Integration

• GitHub account connection support for sandbox gh CLI.
• On-demand Office document preview support in sandbox environments.

Changed

Startup & Runtime Performance

• Reduced startup latency through optimized service initialization and activation flows.
• Reduced build, packaging, and runtime footprint.

Service Architecture Simplification

• Consolidated host services into a unified bundle.
  • Up to 859 MB smaller payload
  • Up to 5.6× faster warm builds
• Improved packaging reliability and ServiceHub architecture.

UI Modernization

• Continued frontend refactoring and modularization.
• Reduced bundle size and improved maintainability.
• Improved feedback workflows, diagnostics, and chat experience.

Telemetry & Observability

• Expanded telemetry infrastructure and deployment automation.
• Improved consent handling, crash reporting, and startup performance monitoring.

Fixed

Chat & UI Reliability

• Fixed viewport shaking, prompt flickering, and streaming interruptions.
• Preserved streaming output across session switches.
• Fixed session title generation, OAuth polling regressions, and upload synchronization issues.
• Improved voice-input error messaging.

Protected File Processing

• Fixed protected-file routing for cloud model sessions.
• Added file-size limits and improved protection safeguards.
• Improved batch protection stability and preview behavior.

Agent & Model Routing

• Fixed tool-call loops and routing edge cases.
• Improved local agent stability, retry behavior, and ambiguity handling.
• Fixed model switching consistency and runtime recovery.

Installer & Deployment

• Improved startup resilience and self-healing.
• Fixed uninstall cleanup issues.
• Improved packaging reliability and launch behavior.
• Resolved model recovery and startup race conditions.

MCP & Plugin Reliability

• Fixed permission and authorization issues.
• Improved tool governance and timeout handling.
• Filtered non-connected plugins and removed broken integrations.

Security

• Addressed penetration-test findings and infrastructure hardening recommendations.
• Enforced TLS certificate validation for model downloads.
• Added prompt-injection protections and secret redaction safeguards.
• Strengthened protected-data processing and privacy controls.
• Hardened observability, deployment, authentication, and network isolation.
• Updated vulnerable dependencies and resolved Trivy and Coverity findings.
• Improved token protection, credential validation, and model-download integrity.
• Removed risky file types (.env, .pem, .crt) from skill-package import allowlists.

SuperClaw Release v1.2.0

Highlights

SuperClaw 1.2.0 introduces major improvements across user experience, local AI performance, MCP integrations, security, observability, and enterprise workflows. This release focuses on increasing reliability, simplifying configuration, strengthening protected-file handling, and improving the overall agent experience.

Added

• MCP (Model Context Protocol) support enhancements with improved connection management, permissions, and tool routing.
• Lightweight local OCR tool and skill for on-device document processing.
• Telemetry and observability framework to support performance monitoring and diagnostics.
• Workspace file preview capabilities with support for Office document rendering.
• Safe upgrade support through superclaw-ctl.
• Installer GPU driver compatibility checks and improved hardware validation.
• Startup port-conflict detection with actionable user guidance.
• Cleanup utility (clean-superclaw.ps1) for resetting local SuperClaw environments.
• Local model quality improvements, including support for Qwen MTP-based assets.

Changed

• Unified model provider configuration across Settings for a simpler model management experience.
• Refreshed UI with improved animations, scrolling behavior, diagnostics organization, tooltips, and accessibility.
• Enhanced privacy and telemetry settings for clearer user control and transparency.
• Improved protected-file workflows with better dataset inspection, masking, and user guidance.
• Updated routing and agent policies to better balance local and cloud model usage.
• Expanded installer experience with clearer messaging, hardware checks, and setup validation.
• Improved email agent support, including broader mailbox coverage and privacy flow refinements.
• Enhanced benchmark, performance measurement, and agent evaluation tooling.

Fixed

• Resolved multiple web search reliability and tool-routing issues.
• Fixed agent loops, runaway executions, and tool-calling budget management problems.
• Improved Hybrid Deep Research stability on weaker local models.
• Fixed MCP connection/disconnection synchronization and permission refresh issues.
• Corrected cloud model validation, selection, and error-reporting behavior.
• Addressed sandbox startup, recovery, and runtime stability issues.
• Fixed confidential/sensitive file processing, preview rendering, and workflow routing issues.
• Improved protected-file agent reliability and masking support for additional file types.
• Fixed email agent filtering, Gmail integration, and automated-message handling.
• Resolved numerous UI consistency, state preservation, preview, and navigation defects.
• Fixed installer, packaging, Docker, and build pipeline reliability issues.

Security

• Hardened protected-file processing pipeline with stricter isolation boundaries and enforced safeguards.
• Improved PII protection workflows and sensitive file handling.
• Added validation for Slack credentials and external integration configuration.
• Strengthened email-agent privacy controls and security workflows.
• Removed GPLv3 libreadline dependency from packaged components.
• Added additional runtime safeguards to prevent invalid agent actions and unintended tool invocation.
• Improved backend image verification and startup integrity checks.
• Multiple Coverity, security hardening, and dependency-related fixes.

Performance & Reliability Highlights

• Improved local-agent execution quality and efficiency.
• Enhanced llmrouter token accounting, budgeting, and model attribution.
• Better sandbox recovery and runtime resilience.
• Optimized web search execution and agent-driven workflows.
• Improved startup diagnostics and troubleshooting support.

Due to the end of support for the current version, we recommend downloading and upgrading to the latest available version to continue receiving security updates, bug fixes, and technical support.

SuperClaw Release v1.1.0

🚀 Highlights

• New ServiceHub + LLMRouter architecture for improved reliability, scalability, and maintainability.
• Enhanced local AI experience with restored llama-server support, local 35B model bundles, HDR capabilities, and performance optimizations.
• Added GitHub Copilot and ChatGPT (Codex) OAuth integrations.
• Stronger data protection and security with Protected File Pipeline enhancements and PII detection.
• Improved chat experience with message editing, retry/regenerate, prompt navigation, searchable model selection, and update notifications.
• Significant improvements to installer reliability, local model management, observability, and startup performance.

---

Added

🤖 AI & Model Management

• Restored local llama-server support with enhanced workload routing and model management.
• Added support for local 35B model bundles with HDR capabilities.
• Added local model hardware preflight validation.
• Added workload-aware routing and improved Auto Route behavior.
• Added local context-size override controls.

☁️ Cloud Provider Integrations

• Added GitHub Copilot OAuth integration.
• Added ChatGPT (Codex subscription) OAuth support.
• Improved Gmail OAuth onboarding with automatic browser launch.

💬 User Experience

• Added application update notifications.
• Added inline editing for the latest user message.
• Added retry and in-place response regeneration.
• Added floating user prompt navigation.
• Added searchable model picker.
• Added warnings before downloading experimental models.

🔒 Security & Data Protection

• Added host-managed Protected File Pipeline.
• Integrated NER and LLM-based sensitive data detection.
• Added additional file-protection and terminal-guard capabilities.

📊 Observability & Diagnostics

• Added startup lifecycle logging.
• Added task success telemetry.
• Added enhanced debugging and observability infrastructure.
• Added debug-only bug reporting with log collection support.

---

Changed

🏗️ Architecture

• Introduced the new ServiceHub, Sandbox Manager, and LLMRouter Manager architecture.
• Refactored backend architecture into host and sandbox components.
• Improved orchestration, service reuse, and startup management.

📦 Installer & Setup

• Enhanced installer reliability and upgrade experience.
• Added virtualization checks and reboot notifications.
• Improved cleanup and uninstall behavior.
• Optimized installer compression and packaging performance.

🖥️ Local Model Experience

• Improved local model download progress reporting.
• Added model deletion confirmation dialogs.
• Simplified provider

Due to the end of support for the current version, we recommend downloading and upgrading to the latest available version to continue receiving security updates, bug fixes, and technical support.

SuperClaw Release v1.0.0

SuperClaw is an AI-powered desktop platform that enables local, cloud, and hybrid AI workflows while providing agent-based automation, research, coding assistance, document processing, and productivity capabilities in a unified experience.

Highlights

🤖 AI Agents and Automation

• Built-in AI agents for research, coding, email assistance, file management, and productivity workflows.
• Multi-agent orchestration framework with task delegation and tool execution capabilities.
• Support for Deep Research workflows, including hybrid and local research experiences.
• Advanced intent classification and intelligent task routing.

🧠 Flexible AI Model Platform

• Support for local, cloud, edge, and auto-routed AI models.
• Unified LLM Router for intelligent model selection and routing.
• Runtime configuration of models and providers through the user interface.
• Support for multiple model providers and cloud AI services.

📧 Email Agent

• Integrated Email Agent with support for drafting, replying, and scheduling actions.
• Credential management and security controls for email workflows.
• Confirmation gates for critical actions such as send, reply, and meeting scheduling.

📁 Workspace and File Intelligence

• Workspace-centric experience for managing projects and files.
• File upload, preview, parsing, and document processing capabilities.
• Support for PDFs, Office documents, and large project repositories.
• Local file analysis and document-aware agent workflows.

🔍 Research and Knowledge Discovery

• Deep Research capabilities for gathering, organizing, and synthesizing information.
• Web-enabled research workflows.
• Document parsing and information extraction tools.
• Search and knowledge exploration across multiple sources.

🛡️ Security and Privacy

• Integrated privacy and protected-file workflows.
• Security hardening across desktop, backend, and dependency ecosystem.
• Vulnerability remediation, dependency updates, and release security reviews.
• Improved safeguards around prompts, URLs, file access, and agent operations.

🎨 User Experience

• Guided onboarding experience and first-run setup.
• Modernized dashboards, session management, and workflow navigation.

🌍 Deployment and Operations

• Windows installer with automated environment setup.
• Linux support through the superclaw-ctl management CLI.
• Edge server connectivity and remote model deployment scenarios.
• CI/CD pipelines, testing infrastructure, release automation, and packaging improvements.

Due to the end of support for the current version, we recommend downloading and upgrading to the latest available version to continue receiving security updates, bug fixes, and technical support.