aiden: an autonomous work engine that plans, executes, and verifies multi-step computer tasks across system interfaces

What it solves

Aiden is an autonomous AI engine designed to move beyond simple AI advice and perform real computer work. It solves the problem of fragmented, multi-step tasks by allowing a user to provide a high-level goal, which the agent then plans, executes, and verifies across various system interfaces.

How it works

Aiden operates as a loop of Goal $\rightarrow$ Plan $\rightarrow$ Tools $\rightarrow$ Permission $\rightarrow$ Execution $\rightarrow$ Recovery $\rightarrow$ Evidence $\rightarrow$ Result. It integrates with the local system (files, terminal, browser) and external services (APIs, MCP servers) to perform actions. It features a tiered approval system to ensure the user remains in control of consequential actions and a verification layer that checks claims against real evidence to ensure the work was actually completed.

Who it’s for

It is designed for developers and power users who need to automate complex workflows, such as codebase maintenance, deep research, browser-based automation, and system operations, across Windows, Linux, WSL, and macOS (via API mode).

Highlights

• Extensive Toolset: Includes 76 skills and 121 tools across 19 providers and 9 channels.
• Durable Autonomy: Tasks are tracked via job-cards, allowing the agent to resume work after restarts and retry recoverable failures.
• Browser Integration: Supports accessibility-tree snapshots, multi-tab handling, and visual screenshot-to-vision capabilities.
• MCP Support: Acts as a Model Context Protocol client to connect to external servers.
• Control Mechanisms: Features a "trust dial" (Observer, Assistant, Partner) and risk-tiered approvals to manage safety and spending.
• Multi-Agent Capability: Can delegate specific parts of a larger job to isolated sub-agents.
• Autonomous Triggers: Can react to files, schedules, email, and webhooks via an optional daemon mode.