* feat(mcp): add per-project MCP server configuration - Add mcpServers config to ProjectEnvConfig type for per-project overrides - Update env-handlers to read/write MCP config from .auto-claude/.env - Update backend get_required_mcp_servers() to respect project config - Refactor AgentTools.tsx to show project-specific MCP toggles - Move MCP Overview to Project section in sidebar navigation - Add i18n translations for MCP server names and descriptions - Update tests for new mcp_config parameter behavior Users can now enable/disable Context7, Linear, Electron, and Puppeteer MCP servers on a per-project basis. Settings are stored in each project's .auto-claude/.env file and respected by the backend when starting agents. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(frontend): send final plan state before unwatching on task exit The file watcher was being stopped before the final plan state could be sent to the renderer. This caused tasks to show stale data (0/0 subtasks) in the UI when they had actually completed successfully with subtasks. Now the final plan is sent to the renderer before unwatching, ensuring the UI receives the correct subtask count and completion status. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * ci(beta-release): add Flatpak packaging support for Linux builds The Linux build was failing with "spawn flatpak ENOENT" because the beta-release workflow was missing the Flatpak setup step that was added to the main release workflow in #404. Adds: - Setup Flatpak step with flatpak-builder and required runtimes - .flatpak to artifact uploads and validation 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(ui): make kanban columns responsive to available width Fixed-width columns wasted horizontal space on wider displays and unnecessarily truncated task titles. Columns now grow with flex-1 while respecting min/max bounds (288-480px for tasks, 320-512px for roadmap). Badge area also expanded slightly (160→180px) to accommodate wider cards. * feat(settings): add user-configurable utility agent settings Make merge_resolver and commit_message agents configurable via the new "Utility" feature setting in Agent Settings. Previously these were hardcoded to Haiku with low thinking, but now users can select their preferred model and thinking level. Changes: - Add utility feature key to FeatureModelConfig/FeatureThinkingConfig - Update AgentTools.tsx to use feature settings instead of fixed - Pass UTILITY_MODEL_ID and UTILITY_THINKING_BUDGET env vars to backend - Backend merge_resolver and commit_message read from env vars - Add i18n translations for utility settings 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: remove unused agent-tools entry from sidebar navigation This commit cleans up the Sidebar component by removing the 'agent-tools' entry from the tools navigation items, streamlining the user interface. The 'worktrees' entry remains intact, ensuring continued access to relevant features. * fix(robustness): address PR review findings for error handling and validation Fix 9 issues identified in PR #424 review: Medium issues: - Add try/except for int() conversion of UTILITY_THINKING_BUDGET env var - Pass '0' when thinking level is 'none' to properly disable extended thinking - Add error handling (|| exit 1) for Flatpak install commands in CI Low issues: - Log exceptions in load_project_mcp_config instead of silent pass - Handle None input in _map_mcp_server_name to prevent AttributeError - Cast mcp_config values to string before split() to handle non-string values - Filter effectiveMcps by project-level MCP states in AgentTools - Log JSON parse errors in getUtilitySettings for easier debugging 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * chore(ci): remove CLA workflow (using hosted CLA Assistant) The CLA workflow was accidentally re-added by PR #254. We use the hosted CLA Assistant service (cla-assistant.io) which handles CLA signing via GitHub webhooks, so this workflow file is redundant and causes failing checks. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(ui): correct graphiti-memory server name in MCP filter The switch case incorrectly used 'graphiti' instead of 'graphiti-memory' which is the actual server ID used throughout the codebase. This caused the filter to not properly check the graphiti MCP server enabled state. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(utility): correctly disable extended thinking when set to "none" When utility thinking level is set to "none", the frontend was sending '0' to the backend, which parsed it as integer 0. The SDK expects max_thinking_tokens=None to disable extended thinking, not 0. Frontend now sends empty string for disabled thinking, and backend interprets empty string as None instead of falling back to 1024. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix issue with github PR checking bot detection * feat(ui): add Claude Code CLI detection and one-click installation Adds comprehensive Claude Code CLI integration to the frontend: - New onboarding step to check if Claude Code is installed - Persistent status badge in sidebar showing version status - Version checking against npm registry with 24h cache - One-click install/update using user's preferred terminal - Cross-platform support (macOS, Windows, Linux) with 20+ terminals - Warning dialog before updates to prevent data loss from killed sessions - Automatic detection of running Claude processes with graceful termination - Added ~/.local/bin to macOS PATH search for Claude CLI detection - Full i18n support (English and French translations) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(ideation): close panel on dismiss and scope events by project Two bugs fixed based on user feedback: 1. Dismiss idea panel not closing: The detail panel now calls onClose() after dismissing, so it closes automatically instead of staying open with hidden action buttons. 2. Idea regeneration affecting all projects: Added currentProjectId tracking to the ideation store. All IPC listeners now filter events by projectId, preventing cross-project state contamination when multiple projects are open. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(github): add parallel orchestrator for PR reviews Implement AI-orchestrated parallel review system using Claude Agent SDK subagents for both initial and follow-up PR reviews. Initial review uses 5 specialist agents: - security-reviewer: OWASP Top 10, injection, auth issues - quality-reviewer: complexity, duplication, error handling - logic-reviewer: algorithm correctness, edge cases, race conditions - codebase-fit-reviewer: naming conventions, pattern adherence - ai-triage-reviewer: validate CodeRabbit, Cursor, Gemini comments Follow-up review uses 3 specialist agents: - resolution-verifier: AI-powered verification of previous findings - new-code-reviewer: security/logic/quality checks on new code - comment-analyzer: triage contributor and AI bot feedback Key features: - AI decides which agents to invoke (not programmatic rules) - User-configurable models via frontend settings (no hardcoding) - SDK handles parallel execution automatically - Cross-validation boosts confidence when agents agree Also fixes bot_detection.py import error for relative imports. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(core): add agents parameter to create_client for SDK subagents The create_client function was missing support for the `agents` parameter needed by the parallel orchestrator reviewers to define SDK subagents. This enables the parallel PR review system to define specialist agents (resolution-verifier, new-code-reviewer, comment-analyzer) that the SDK can execute in parallel. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(security): add usedforsecurity=False to MD5 hash for finding IDs The MD5 hash is used for generating unique finding IDs (non-security purpose), so Bandit B324 warning is addressed by marking it explicitly. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(github): add PR review logs feature and parallel orchestrator improvements - Add PR logs feature to view AI review thinking/tool usage during analysis - Create PRLogCollector class to capture and structure subprocess output - Add PRLogs component with collapsible phases (context, analysis, synthesis) - Improve parallel orchestrator and followup reviewer with better agent coordination - Add bot detection improvements and fix benefit-of-doubt logic - Add comprehensive tests for PR review, bot detection, and E2E flows - Add IPC channel and browser mock for PR logs retrieval - Add i18n translations for review logs UI 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(github): show PR review logs during AI analysis in progress - Add logs section that appears when review is in progress, not just after completion - Add periodic log refresh (2s interval) while review is streaming - Add isStreaming prop to PRLogs component for live indicator - Show "Live" badge on logs header during active review - Show streaming status on active phases 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(github): show actual AI response content in PR review logs - Update Python orchestrators to print AI response text preview (up to 500 chars) - Filter out unhelpful debug messages like "Message #15: AssistantMessage" - Add ParallelOrchestrator to log source patterns and color mapping - Move Followup to analysis phase (not context) for better categorization The logs now show actual AI thinking and responses instead of just message types. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(github): capture synthesis phase logs and mark all phases complete - Add parsing for [PR Review Engine], [PR #XXX] progress, and Summary lines - Add PR progress message pattern matching for [PR #XXX] [YY%] format - Map PR Review Engine, Summary, and Progress sources to synthesis phase - Update finalize() to mark pending phases as completed when review succeeds - Add source colors for PR Review Engine (indigo) and Summary (emerald) The synthesis phase now properly shows logs and marks as Complete. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(ui): merge tools section into project and add dynamic nav filtering Moves GitHub Issues, GitHub PRs, GitLab Issues, and GitLab MRs tabs from the separate TOOLS section into the PROJECT section. Navigation items are now dynamically filtered based on project settings - GitHub tabs only show when GitHub is enabled, and GitLab tabs only show when GitLab is enabled. This reduces visual clutter by hiding integrations that aren't configured while consolidating all project-related navigation into a single section. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(pr-review): implement strict quality gates severity system Redesign PR review severity labels and verdict logic based on research: - CRITICAL → "Blocker" (blocks merge) - HIGH → "Required" (blocks merge) - MEDIUM → "Recommended" (blocks merge - AI fixes quickly) - LOW → "Suggestion" (optional) Key changes: - Medium severity findings now result in NEEDS_REVISION verdict - Only LOW severity allows MERGE_WITH_CHANGES - Updated all 5 verdict logic files for consistency - Updated AI prompts with strict quality gates guidance - Updated en/fr i18n labels with action-oriented terminology Rationale: AI can fix code issues quickly, so be aggressive about code quality. 95% of fixes are done by AI anyway. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(mcp): add health checks and bearer token auth for custom MCP servers Users adding HTTP MCP servers had no way to know if the server was healthy, needed authentication, or was unreachable. This adds: - Health status indicators (healthy/needs auth/unhealthy/checking) - Quick connectivity check on component mount - Manual "Test" button for full MCP protocol test - Simple "Authentication Token" field that creates Bearer header - URL pattern detection with helpful hints for known providers (GitHub, Google, Anthropic, OpenAI) with links to create tokens - Collapsible "Advanced Headers" section for custom headers - i18n translations for all new fields (EN/FR) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * feat(gitlab): add glab CLI detection and one-click install When users try to use OAuth for GitLab authentication, the app now checks if glab CLI is installed first. If not installed, it shows an inline warning card with a one-click install button that opens the user's preferred terminal with the appropriate install command (brew for macOS, winget for Windows, snap/brew for Linux). This prevents the silent failure that occurred when glab was missing, where the OAuth flow would fail with ENOENT and leave users with a perpetual loading spinner. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(pr-review): pass USE_CLAUDE_MD env var to subprocess The PR review subprocess wasn't receiving the project's useClaudeMd setting, causing it to always show "CLAUDE.md: disabled by project settings" even when enabled in the UI. Changes: - Added optional `env` parameter to SubprocessOptions interface - Updated runPythonSubprocess to merge custom env vars with filtered env - PR handlers now pass USE_CLAUDE_MD based on project.settings.useClaudeMd 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(pr-review): improve agent invocation and findings logging The logs previously showed "Agents invoked: []" even when agents were running because the streaming detection wasn't reliable. Also, the findings summary was hidden. Changes: - Extract agents from structured output (reliable source) instead of streaming detection which was returning empty - Log each specialist agent with [Agent:name] label when complete - Add detailed findings summary showing severity, title, file:line - Both parallel orchestrator and followup reviewer updated Example new log output: [ParallelOrchestrator] Specialist agents invoked: security-reviewer, logic-reviewer [Agent:security-reviewer] Analysis complete [Agent:logic-reviewer] Analysis complete [ParallelOrchestrator] Findings summary: [LOW] 1. Suggestion title (file.ts:42) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor(pr-review): enhance quality gates and logging for PR assessments Updated the PR review process to enforce stricter quality gates for severity levels, ensuring that HIGH and MEDIUM issues block merges. Adjusted the verdict criteria for clarity and consistency across the system. Enhanced logging for PR reviews to provide real-time updates and improved user feedback. Changes: - Revised verdict criteria to reflect strict quality gates - Updated logging to capture all findings, including LOW severity suggestions - Improved real-time log streaming during PR reviews This ensures a more robust and user-friendly review process, emphasizing the importance of addressing all findings. 🤖 Generated with [Claude Code](https://claude.com/claude-code) * feat(pr-review): add real-time log streaming and specialist agent tracking - Add incremental log saving in PRLogCollector (every 3 entries) for real-time streaming - Add phase transition tracking to properly mark phases as complete - Add parsing for specialist agent logs ([Agent:xxx] format) - Add color-coded badges for specialist agents in frontend logs UI - Track subagent invocations via ToolUseBlock/ToolResultBlock in message content 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(pr-review): improve verdict display, follow-up timing, and findings UX Three fixes for PR review: 1. Verdict message now includes LOW findings count (e.g., "1 required, 0 recommended, 4 suggestions") 2. "Ready for Follow-up" only shows when commits happen AFTER findings are posted, not during/before the review 3. Posted findings are hidden from selection UI - shows "All findings posted to GitHub" instead of confusing "0/5 selected" Also removes legacy orchestrator_reviewer.py (dead code superseded by parallel orchestrator) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(review): address PR #424 code review feedback Fixes issues flagged by CodeRabbit, Cursor bot, and GitHub security: - Fix nullish coalescing for 'none' thinking budget (worktree-handlers.ts) - Add set -e to Flatpak CI setup for consistent error handling - Add explicit UTF-8 encoding to file open in client.py - Add type="button" to 10 buttons to prevent form submissions - Remove unused isLoading and getServerStatus variables - Improve French translations with proper definite articles 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(pr-review): ensure synthesis tab shows completed status after review When a follow-up review completed, the Synthesis tab would continue showing "Running" instead of "Complete". This was due to a race condition where the log polling stopped immediately when isReviewing became false, before fetching the final log state with completed phase statuses. Added a final log refresh when the review completes by tracking the previous isReviewing state and fetching logs one more time when the state transitions from true to false. * fix(security): address code review security and quality issues Fixes security vulnerabilities and code quality issues from Auto Claude review: - Fix command injection: use execFileSync with args array instead of template string interpolation for git commands (worktree-handlers.ts) - Fix JSON injection: add schema validation for CUSTOM_MCP_SERVERS to reject malicious configurations (client.py) - Fix code smell: use proper destructuring for unused state variable - Add i18n: replace hardcoded English strings with translation keys - Extract shared utility: create core/model_config.py to eliminate duplicate model/thinking budget parsing code (DRY violation) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(pr-review): clear stale logs when starting new follow-up review When starting a second follow-up review, the UI was showing the previous review's completed phase statuses instead of fresh pending states. This happened because the logs state wasn't cleared when a new review started. Now clears the logs state when isReviewing transitions from false to true, ensuring fresh logs are fetched and displayed. * fix(security): address remaining command injection vulnerabilities Fixes HIGH and MEDIUM severity issues from PR review: Security fixes: - Remove shell: true from spawn() in mcp-handlers.ts checkCommandHealth and testCommandConnection to prevent shell metacharacter injection - Add command allowlist (npx, npm, node, python) and blocklist (bash, sh, cmd, powershell) to _validate_custom_mcp_server() in client.py - Fix type validation mismatch: 'url' -> 'http' to match downstream usage Quality fixes: - Add negative value validation for UTILITY_THINKING_BUDGET in model_config.py - Replace silent error swallowing with console.error in PRDetail.tsx 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor(pr-review): extract shared utilities and reduce complexity - Extract SDK stream processing into sdk_utils.py (~374 lines removed) - Callback-based architecture for message/thinking/error handling - Eliminates duplicate stream processing in 3+ reviewer modules - Extract category mapping into category_utils.py (~80 lines removed) - Unified CATEGORY_MAPPING dictionary - Single source of truth for severity/category translations - Refactor parallel_orchestrator_reviewer.py review() method - Split 380-line method into 165 lines + 8 focused helpers - Each extracted method under 50 lines for maintainability - Extracted: _prepare_context, _create_specialist_inputs, etc. - Remove unused ReviewCategory imports after extraction Total: ~454 lines of duplicate code eliminated 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(security): address all remaining PR #424 review findings Backend security hardening (client.py): - Reject commands with path separators (/ or \) to prevent path traversal - Add dangerous interpreter flags blocklist (--eval, -e, -c, --exec) - Add pwsh (PowerShell Core) to DANGEROUS_COMMANDS Frontend defense-in-depth (mcp-handlers.ts): - Add SAFE_COMMANDS allowlist with path validation before spawn - Add OS-level timeout (15000ms) to testCommandConnection spawn Model config fix: - Treat UTILITY_THINKING_BUDGET=0 as "disable thinking" (same as empty) SDK stream processing improvements (sdk_utils.py): - Add try/except for stream-level and message-level errors - Add warning when multiple StructuredOutput blocks overwrite previous - Return error field in result dict for caller visibility Code consolidation: - Remove duplicate _CATEGORY_MAPPING from review_tools.py, use shared module - Remove unreachable 'best-practices' entry in category_utils.py 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(pr-review): capture full summary content in synthesis logs Extended the log parsing patterns to capture markdown content that appears in the review summary. Previously, only header lines like "Summary:" were captured, but the actual content (markdown headers, bullet points, numbered lists, findings, file references) was discarded because it didn't match any pattern. Added patterns for: - Markdown headers (##, ###) - Bullet points and indented findings - Bold text lines - Numbered lists - File references - Additional summary fields (Is Follow-up, Resolved, etc.) * fix(pr-review): sync PR list status with detail view using overallStatus The PR list was computing status based only on HIGH/CRITICAL severity findings, while the PR detail used the overallStatus field from the backend. This caused inconsistent display where list showed "Ready to Merge" but detail showed "Changes Requested" for MEDIUM severity issues. Fixed by using overallStatus as the source of truth in both: - PRList.tsx: hasBlockingFindings prop computation - usePRFiltering.ts: getPRComputedStatus function * fix(security): address follow-up review findings round 2 Backend security (client.py): - Expand DANGEROUS_FLAGS to include: -m (Python module), -p (Python eval+print), --print, --input-type=module, --experimental-loader, --require, -r Frontend defense-in-depth (mcp-handlers.ts): - Add DANGEROUS_FLAGS set mirroring backend - Add areArgsSafe() function to validate args - Check args in both checkCommandHealth and testCommandConnection before spawn SDK stream error handling: - Add error field check in parallel_orchestrator_reviewer.py - Add error field check in parallel_followup_reviewer.py - Raise RuntimeError on stream failure instead of silently continuing Model config: - Add debug log when UTILITY_THINKING_BUDGET=0 disables thinking 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Auto Claude
Autonomous multi-agent coding framework that plans, builds, and validates software for you.
Download
Stable Release
| Platform | Download |
|---|---|
| Windows | Auto-Claude-2.7.1-win32-x64.exe |
| macOS (Apple Silicon) | Auto-Claude-2.7.1-darwin-arm64.dmg |
| macOS (Intel) | Auto-Claude-2.7.1-darwin-x64.dmg |
| Linux | Auto-Claude-2.7.1-linux-x86_64.AppImage |
| Linux (Debian) | Auto-Claude-2.7.1-linux-amd64.deb |
Beta Release
⚠️ Beta releases may contain bugs and breaking changes. View all releases
| Platform | Download |
|---|---|
| Windows | Auto-Claude-2.7.2-beta.10-win32-x64.exe |
| macOS (Apple Silicon) | Auto-Claude-2.7.2-beta.10-darwin-arm64.dmg |
| macOS (Intel) | Auto-Claude-2.7.2-beta.10-darwin-x64.dmg |
| Linux | Auto-Claude-2.7.2-beta.10-linux-x86_64.AppImage |
| Linux (Debian) | Auto-Claude-2.7.2-beta.10-linux-amd64.deb |
| Linux (Flatpak) | Auto-Claude-2.7.2-beta.10-linux-x86_64.flatpak |
All releases include SHA256 checksums and VirusTotal scan results for security verification.
Requirements
- Claude Pro/Max subscription - Get one here
- Claude Code CLI -
npm install -g @anthropic-ai/claude-code - Git repository - Your project must be initialized as a git repo
- Python 3.12+ - Required for the backend and Memory Layer
Quick Start
- Download and install the app for your platform
- Open your project - Select a git repository folder
- Connect Claude - The app will guide you through OAuth setup
- Create a task - Describe what you want to build
- Watch it work - Agents plan, code, and validate autonomously
Features
| Feature | Description |
|---|---|
| Autonomous Tasks | Describe your goal; agents handle planning, implementation, and validation |
| Parallel Execution | Run multiple builds simultaneously with up to 12 agent terminals |
| Isolated Workspaces | All changes happen in git worktrees - your main branch stays safe |
| Self-Validating QA | Built-in quality assurance loop catches issues before you review |
| AI-Powered Merge | Automatic conflict resolution when integrating back to main |
| Memory Layer | Agents retain insights across sessions for smarter builds |
| GitHub/GitLab Integration | Import issues, investigate with AI, create merge requests |
| Linear Integration | Sync tasks with Linear for team progress tracking |
| Cross-Platform | Native desktop apps for Windows, macOS, and Linux |
| Auto-Updates | App updates automatically when new versions are released |
Interface
Kanban Board
Visual task management from planning through completion. Create tasks and monitor agent progress in real-time.
Agent Terminals
AI-powered terminals with one-click task context injection. Spawn multiple agents for parallel work.
Roadmap
AI-assisted feature planning with competitor analysis and audience targeting.
Additional Features
- Insights - Chat interface for exploring your codebase
- Ideation - Discover improvements, performance issues, and vulnerabilities
- Changelog - Generate release notes from completed tasks
Project Structure
Auto-Claude/
├── apps/
│ ├── backend/ # Python agents, specs, QA pipeline
│ └── frontend/ # Electron desktop application
├── guides/ # Additional documentation
├── tests/ # Test suite
└── scripts/ # Build utilities
CLI Usage
For headless operation, CI/CD integration, or terminal-only workflows:
cd apps/backend
# Create a spec interactively
python spec_runner.py --interactive
# Run autonomous build
python run.py --spec 001
# Review and merge
python run.py --spec 001 --review
python run.py --spec 001 --merge
See guides/CLI-USAGE.md for complete CLI documentation.
Configuration
Create apps/backend/.env from the example:
cp apps/backend/.env.example apps/backend/.env
| Variable | Required | Description |
|---|---|---|
CLAUDE_CODE_OAUTH_TOKEN |
Yes | OAuth token from claude setup-token |
GRAPHITI_ENABLED |
No | Enable Memory Layer for cross-session context |
AUTO_BUILD_MODEL |
No | Override the default Claude model |
GITLAB_TOKEN |
No | GitLab Personal Access Token for GitLab integration |
GITLAB_INSTANCE_URL |
No | GitLab instance URL (defaults to gitlab.com) |
LINEAR_API_KEY |
No | Linear API key for task sync |
Building from Source
For contributors and development:
# Clone the repository
git clone https://github.com/AndyMik90/Auto-Claude.git
cd Auto-Claude
# Install all dependencies
npm run install:all
# Run in development mode
npm run dev
# Or build and run
npm start
System requirements for building:
- Node.js 24+
- Python 3.12+
- npm 10+
Installing dependencies by platform:
Windows
winget install Python.Python.3.12
winget install OpenJS.NodeJS.LTS
macOS
brew install python@3.12 node@24
Linux (Ubuntu/Debian)
sudo apt install python3.12 python3.12-venv
curl -fsSL https://deb.nodesource.com/setup_24.x | sudo -E bash -
sudo apt install -y nodejs
Linux (Fedora)
sudo dnf install python3.12 nodejs npm
See CONTRIBUTING.md for detailed development setup.
Building Flatpak
To build the Flatpak package, you need additional dependencies:
# Fedora/RHEL
sudo dnf install flatpak-builder
# Ubuntu/Debian
sudo apt install flatpak-builder
# Install required Flatpak runtimes
flatpak install flathub org.freedesktop.Platform//25.08 org.freedesktop.Sdk//25.08
flatpak install flathub org.electronjs.Electron2.BaseApp//25.08
# Build the Flatpak
cd apps/frontend
npm run package:flatpak
The Flatpak will be created in apps/frontend/dist/.
Security
Auto Claude uses a three-layer security model:
- OS Sandbox - Bash commands run in isolation
- Filesystem Restrictions - Operations limited to project directory
- Dynamic Command Allowlist - Only approved commands based on detected project stack
All releases are:
- Scanned with VirusTotal before publishing
- Include SHA256 checksums for verification
- Code-signed where applicable (macOS)
Available Scripts
| Command | Description |
|---|---|
npm run install:all |
Install backend and frontend dependencies |
npm start |
Build and run the desktop app |
npm run dev |
Run in development mode with hot reload |
npm run package |
Package for current platform |
npm run package:mac |
Package for macOS |
npm run package:win |
Package for Windows |
npm run package:linux |
Package for Linux |
npm run package:flatpak |
Package as Flatpak |
npm run lint |
Run linter |
npm test |
Run frontend tests |
npm run test:backend |
Run backend tests |
Contributing
We welcome contributions! Please read CONTRIBUTING.md for:
- Development setup instructions
- Code style guidelines
- Testing requirements
- Pull request process
Community
- Discord - Join our community
- Issues - Report bugs or request features
- Discussions - Ask questions
License
AGPL-3.0 - GNU Affero General Public License v3.0
Auto Claude is free to use. If you modify and distribute it, or run it as a service, your code must also be open source under AGPL-3.0.
Commercial licensing available for closed-source use cases.


