What Is Dorabot and Why Did It Launch Now?
Dorabot is an open-source macOS application that transforms coding agents like Claude Code or Codex into autonomous, always-on AI agents that operate continuously without constant supervision. It launched because developers are exhausted by the stop-and-start nature of terminal-based coding assistants that lose context between sessions and require human initiation for every task. The app provides a native desktop harness featuring persistent memory, workspace management, and proactive scheduling through a heartbeat system. Unlike reactive chat interfaces that wait for your prompt, Dorabot uses OpenClaw-inspired pulse mechanisms to wake up on schedules, scan for work, and propose actions. It runs entirely locally with no cloud relay, ships under the MIT license, and was built by the developer suitedaces to solve the specific problem of automating competitive research, email management, and coding workflows while maintaining complete privacy and control over your data. This approach addresses a significant gap in the current AI agent landscape, providing a robust, user-controlled solution for continuous operation.
How Does Dorabot Transform Claude Code Into a 24/7 Agent?
Claude Code delivers powerful terminal-based coding assistance but stops when you close the window. Dorabot wraps this capability in a persistent macOS process that keeps your agent alive indefinitely. The transformation occurs through sophisticated process management and state persistence that maintains context across days and weeks. Instead of starting fresh with each chat session, Dorabot creates a dedicated workspace where the agent stores research notes, daily journals, task lists, and code context. The app handles process lifecycle management, automatically restarting components if they crash and providing a heartbeat mechanism that triggers autonomous behavior. This architectural shift means Claude Code can monitor your GitHub issues, triage emails, run continuous integration tests, or perform market research while you sleep. The native desktop UI displays real-time status, pending approvals requiring your attention, and searchable memory contents, effectively converting a reactive tool into a proactive teammate. This continuous operation enables a new level of automation for developers.
What Makes Dorabot’s Heartbeat System Different?
Most AI agents remain passive until you type a command. Dorabot’s heartbeat system, directly inspired by OpenClaw architecture, fundamentally inverts this relationship through scheduled autonomy. The system uses cron-like intervals or custom event triggers to wake the agent periodically regardless of your presence. When a heartbeat fires, the agent scans its persistent memory, checks configured data sources including email servers, calendars, and RSS feeds, then identifies actionable items requiring attention. It composes task proposals that surface through the desktop UI or push to your messaging apps, waiting for explicit approval before execution. This creates genuine proactive autonomy rather than reactive assistance. You configure heartbeat intervals in the workspace YAML file, setting it to check cryptocurrency markets every fifteen minutes, review pull requests hourly, or scan competitor websites twice daily. The system prevents agent stagnation and ensures continuous operation without requiring you to remember to initiate tasks manually.
How Does Persistent Memory Work in Dorabot?
Memory in Dorabot extends far beyond simple conversation history into a structured knowledge base utilizing local SQLite databases, JSON stores, and markdown journals. The agent writes daily activity logs, research findings, and task outcomes to disk after every operation. When subsequent heartbeats trigger, the agent reads this accumulated context to inform current decisions and avoid redundant work. The memory system supports semantic search through embeddings, allowing the agent to recall relevant information from weeks prior without explicit prompting. For instance, if you researched a competitor’s pricing strategy three weeks ago, the agent remembers those findings when asked about market positioning today. The workspace directory structure strictly separates your project code from the agent’s cognitive memory, preventing the agent from accidentally committing its internal thoughts to your git repositories. This persistent state architecture enables long-term projects spanning months rather than minutes.
Can You Really Code From WhatsApp and Telegram?
Yes, and this capability fundamentally alters your relationship with development work. Dorabot integrates with messaging platforms through native APIs and webhook configurations that you set up in the application preferences. You provide your Telegram bot token or WhatsApp Business API credentials, and the agent establishes a persistent connection to these services. Once connected, you send natural language commands like “deploy the hotfix to production” or “refactor the authentication module using the new pattern” directly from your phone. The agent processes these requests in the background on your Mac and replies with execution results, diffs, or questions requiring clarification. This proves invaluable when you are away from your desk but need to unblock a team member, restart a failed service, or check build status. The messaging layer respects the same approval gates as the desktop UI, ensuring that destructive operations still require your explicit consent before execution. This remote control capability offers unparalleled flexibility.
How Does Dorabot Handle Browser Automation?
Dorabot utilizes the Chrome DevTools Protocol (CDP) for native browser control rather than wrapping Selenium or Playwright. This approach gives the agent direct access to Chrome’s internals for maximum reliability and speed. The agent can launch headless or headed browser instances, navigate complex single-page applications, scrape structured data, fill multi-step forms, capture screenshots, and download files. Because everything runs locally, your existing Chrome cookies, authentication sessions, and extensions remain available to the agent. The CDP integration enables sophisticated competitive research workflows where the agent logs into dashboards, extracts metrics, and monitors pricing changes. You extend browser capabilities by writing custom CDP scripts or installing MCP servers that expose additional browser tools. The agent maintains browser state in its persistent memory, allowing it to resume multi-step research tasks exactly where it left off after interruptions or system restarts. This direct control minimizes overhead and improves performance.
What Is the Dorabot Workspace Architecture?
The workspace implements a strict directory structure that enforces separation of concerns between your work and the agent’s cognition. Your project code resides in /workspace/user/ while the agent’s memory, journals, configuration files, and skill definitions live in /workspace/agent/. This isolation prevents the agent from accidentally modifying your source code or committing its internal monologue to version control. The architecture supports multiple concurrent workspaces, letting you run separate agent instances for different clients, projects, or domains. Each workspace maintains its own .env file for API secrets, config.yaml for heartbeat scheduling, and skills/ directory containing MCP server binaries. The macOS app provides a built-in file explorer view of this structure, making it trivial to inspect what the agent knows, debug corrupted memory files, or manually prune old journals when storage grows large. This modular design enhances organization and control.
How Extensible Is Dorabot for Custom Skills?
Extensibility centers on the Model Context Protocol (MCP), an open standard that Dorabot implements as a client. You write skills in Python, TypeScript, Go, or any language supporting MCP stdio or HTTP transports. Common extensions include database connectors for PostgreSQL, Slack notification bots, custom REST API clients for internal services, or specialized data processors. The skill system supports hot-reloading; placing a new server binary in the skills directory makes it available on the next heartbeat without restarting the application. You can override built-in behaviors by providing custom system prompts or tool definitions that take precedence over defaults. This architecture ensures Dorabot grows with your specific workflow requirements without requiring you to fork or modify the core Swift codebase, creating an ecosystem of reusable agent capabilities. This allows for significant customization and adaptation to diverse needs.
Dorabot vs Other OpenClaw Implementations: A Comparison
| Feature | Dorabot | OpenClaw CLI | Molinar |
|---|---|---|---|
| Platform | macOS native app | Cross-platform terminal | Web-based SaaS |
| Memory | Local SQLite + files | In-memory/Redis | Cloud vector database |
| Proactive Execution | Built-in heartbeat | Requires external cron | Scheduled cloud tasks |
| Messaging | Native WhatsApp/Telegram | Webhooks only | Email notifications |
| Privacy Model | 100% local, no relay | Configurable | Cloud-dependent |
| UI Type | SwiftUI desktop | Terminal TUI | Browser interface |
| License | MIT | MIT | AGPL |
Dorabot distinguishes itself through its native macOS integration and local-first philosophy compared to terminal-based OpenClaw implementations or web-centric alternatives like Molinar. While OpenClaw CLI offers maximum flexibility for Linux servers, Dorabot provides the polished desktop experience that macOS users expect. Unlike cloud-based solutions, Dorabot keeps your data on-device, eliminating latency and privacy concerns at the cost of requiring your own hardware to remain powered. This table highlights the core differences in approach and functionality, offering a clear view of where Dorabot fits within the broader AI agent landscape. Each implementation caters to different user needs and technical preferences.
What Does Local-Only Architecture Mean for Privacy?
Everything stays on your machine. Zero cloud relay means your proprietary code, email contents, competitive research, and API keys never traverse third-party servers or internet relays. The MIT license guarantees you can audit every line of Swift code handling your data. For developers working with sensitive intellectual property, healthcare data under HIPAA, or regulated financial information, this architecture eliminates data residency concerns and compliance headaches. Your OpenAI or Anthropic API keys reside exclusively in local environment files protected by macOS Keychain or file permissions. Browser automation runs against your local Chrome profile with your actual login sessions rather than isolated cloud browsers. Even messaging integrations route through your local network connection directly to provider APIs, not through a centralized broker service that could log or intercept communications. This commitment to local processing provides a high degree of privacy and security, which is paramount for many users.
How Do You Install and Configure Dorabot?
Installation currently requires manual building from source or downloading precompiled releases from the GitHub repository suitedaces/dorabot. Clone the repository, open the Xcode project, and build for your architecture or download the latest .dmg release. Drag the application to /Applications/ and launch it. On first run, Dorabot creates a default workspace directory at ~/.dorabot/default/. Configuration happens through editing ~/.dorabot/config.yaml and using the native preferences pane. You must add API keys for your chosen LLM provider, define heartbeat intervals using cron syntax, and configure messaging webhooks if desired. The app validates dependencies, checking for Chrome installation for CDP features and Node.js if you plan to run TypeScript-based MCP servers. Initial setup typically requires ten minutes, with the repository providing example configurations for common personas like “autonomous researcher” or “coding assistant with Telegram access.” Detailed instructions and troubleshooting guides are available in the project’s documentation.
What Are the System Requirements for Dorabot?
Dorabot requires macOS 14.0 (Sonoma) or later due to dependencies on modern SwiftUI features and security APIs. You need minimum 8GB RAM to comfortably run both the agent processes and Chrome simultaneously for browser automation. Storage requirements vary by usage but plan for at least 2GB for the application bundle, agent logs, and initial workspace, plus additional space accumulating from memory files and research downloads. Apple Silicon Macs (M1 or newer) are recommended as the primary build targets ARM64 architecture, though Intel Macs can run the application through Rosetta 2 translation with slight performance penalties. Chrome must be installed for CDP browser features to function. Network access is required only for API calls to LLM providers and messaging platform endpoints. The application runs efficiently in the background using minimal CPU when idle, spiking only during heartbeat execution or heavy coding tasks. Ensuring your system meets these specifications will provide the best experience.
How Does Dorabot Compare to Cloud-Based Agent Platforms?
Cloud platforms like AutoGPT Cloud, Devin, or SaaS coding assistants offer convenience but sacrifice control and privacy. Dorabot keeps computation local, meaning you pay only for LLM API tokens rather than per-seat SaaS subscriptions that scale with team size. Latency remains lower because requests route directly from your Mac to Anthropic or OpenAI without intermediate cloud relays adding network hops. However, you lose automatic cross-device synchronization unless you configure your own file syncing solution like Syncthing or Dropbox for the workspace directory. Cloud platforms handle infrastructure maintenance and scaling automatically; Dorabot requires you to keep your Mac running, charged, and online. For solo developers, privacy-conscious startups, or those working with air-gapped sensitive data, the trade-off overwhelmingly favors local control. For large enterprises needing centralized audit logs and team collaboration features, managed cloud solutions might still prove more practical despite the costs. The choice depends on specific priorities.
What MCP Servers Work Best With Dorabot?
Start with the official Model Context Protocol servers for foundational capabilities. The filesystem server provides sandboxed read-write access to your workspace directories. The SQLite server offers structured persistent storage for the agent’s memory and knowledge graphs. The Brave Search server enables web research without requiring CDP browser automation for simple queries. For development workflows, the GitHub MCP server allows the agent to review pull requests, comment on issues, and trigger actions. The Slack and Telegram MCP servers handle bidirectional messaging integration. Community-built servers for PostgreSQL, Redis, and AWS services extend capabilities for infrastructure management. Dorabot’s MCP client supports both stdio and HTTP transports, ensuring compatibility with most community servers written in Python or TypeScript. When installing third-party servers, verify they respect the local-only philosophy by checking they do not phone home to external telemetry services. This ensures continued privacy and security.
How Is Dorabot Changing the Proactive Agent Paradigm?
Reactive agents wait passively for human prompts like sophisticated search engines. Proactive agents initiate actions based on schedules, triggers, or inferred intent. Dorabot makes this paradigm accessible to individual macOS developers without requiring cloud infrastructure or DevOps expertise. The heartbeat system transforms coding agents from tools into autonomous workers that monitor systems, conduct research, and notify you of findings. This shifts your role from operator to supervisor, from initiator to approver. Instead of remembering to check email or run tests, you receive proposals from the agent that has already done the work. The psychological shift is significant: you transition from “I need to research competitors” to “The agent completed the competitive analysis and highlighted three threats requiring your review.” Dorabot demonstrates that open-source, local software can achieve the autonomy previously reserved for expensive enterprise SaaS platforms. This represents a significant advancement in personal AI automation.
What Are the Limitations of Running Dorabot on macOS?
macOS exclusivity immediately excludes teams using Linux workstations or Windows development environments. Background execution faces constraints from Apple’s App Nap and aggressive power management, which may delay heartbeat tasks if the system decides to throttle background processes. There exists no built-in multi-user support; the application targets individual developers rather than shared team environments. Horizontal scaling requires manual effort; running ten agents necessitates either ten separate Macs or complex virtualization, unlike cloud solutions that scale with a configuration change. The SwiftUI interface limits contributions to developers with macOS development experience, potentially slowing community growth compared to JavaScript-based alternatives. Memory files grow unbounded over time unless you manually implement pruning strategies, risking disk space exhaustion on long-running instances. These constraints position Dorabot as ideal for personal productivity and small teams, not enterprise fleet management. Understanding these limitations helps in determining if Dorabot is the right solution for your specific needs.
How Does Dorabot Impact Developer Productivity?
Productivity gains manifest primarily through the elimination of context switching and administrative overhead. By handling email triage, market monitoring, dependency updates, and routine refactoring asynchronously, Dorabot removes the micro-interruptions that fragment deep work sessions. Developers report using the messaging integration to unblock continuous integration pipelines while commuting, review security alerts during lunch, or deploy hotfixes from bed during outages. The persistent memory eliminates the time sink of re-explaining project context to the agent for each new task. However, realize an upfront investment cost in configuring skills, training the agent on your preferences, and establishing trust boundaries for autonomous actions. Once configured, the agent operates at machine speed while you operate at human speed, creating an asymmetric productivity multiplier. The return on investment compounds over weeks as the agent accumulates context and refines its proposal accuracy based on your approval patterns. This long-term benefit makes Dorabot a valuable tool for serious developers.
Where Is Dorabot Headed Next?
The public roadmap includes Linux support through a headless CLI version that maintains feature parity with the macOS GUI, enabling server deployment. Collaborative workspaces for small teams are in development, allowing shared memory and task delegation between multiple human operators. Enhanced memory management using automatic embedding-based pruning will solve the unbounded growth problem. Additional messaging platform support for Discord and iMessage is planned. The community is driving rapid expansion of available MCP servers, with discussions of a curated registry similar to OpenClaw’s tool directory. Future versions may introduce optional encrypted cloud backup for workspace synchronization between devices while maintaining local execution. The core philosophy will remain local-first and MIT-licensed. Track the GitHub repository for weekly releases and breaking changes as the project iterates quickly based on user feedback and contributions from the OpenClaw ecosystem. The future looks promising for Dorabot and its growing capabilities.
Frequently Asked Questions
Is Dorabot free to use?
Yes, Dorabot is completely free and open-source under the MIT license. You can clone the repository from GitHub, build it locally, and use it without subscription fees. You only pay for the API calls to LLM providers like Anthropic or OpenAI based on your token usage.
Can I use Dorabot with OpenAI Codex instead of Claude Code?
Absolutely. Dorabot is model-agnostic and works with any coding agent that exposes a compatible interface. You can configure it to use Claude Code, OpenAI Codex, or even local models via Ollama or LM Studio by adjusting the provider settings in the configuration file.
How secure is the local data storage?
Dorabot stores all data locally on your Mac with no cloud relay or external servers. Your API keys, memory files, and workspace data remain on-device. This architecture ensures that sensitive code, emails, and research data never leave your machine, though you should still encrypt your disk and secure your local environment files.
Does Dorabot work when my Mac is asleep?
No, Dorabot requires your Mac to be awake and running to execute heartbeat tasks. macOS sleep modes will pause execution. To run truly 24/7, you need to disable sleep in System Settings or run Dorabot on a Mac that stays awake, such as a Mac mini server or a MacBook connected to power with sleep disabled.
Can I run multiple Dorabot agents simultaneously?
Yes, you can run multiple agents by creating separate workspace directories for each project or client. Each workspace maintains its own memory, configuration, and heartbeat schedule. However, each instance consumes system resources, so plan your RAM and CPU capacity accordingly when running several agents at once.