Overview

Apache Maka (Incubating) is a groundbreaking local-first AI agent workspace designed for serious work, not just simple queries. It empowers you to build and manage sophisticated AI agents that can interact with your projects, execute tools, and generate tangible outputs. Maka's core innovation lies in its robust, append-only execution log, which meticulously records every model message, tool call, permission decision, and termination event, ensuring complete traceability and enabling powerful recovery capabilities.

Key Features

• Local-First Architecture: Your sessions, settings, and run records stay on your machine by default, offering enhanced privacy and control. You choose how your agent connects to models, whether via cloud APIs, local models, or compatible gateways.
• Runtime Event Log: All agent interactions, including model messages, tool calls, and results, are stored in an immutable log. This log serves as the single source of truth for the agent's execution, enabling reliable recovery and replay.
• Contextual Intelligence: Maka treats context intelligently, allowing for dynamic pruning of tool results and LLM compaction without discarding recorded evidence, ensuring agents operate with efficient yet comprehensive information.
• Unified Execution Authority: The Runtime Host centralizes control over sessions, agent lifecycles, tool execution, and event handling, providing a consistent and secure environment.
• Multiple Interface Surfaces: Interact with Maka through an intuitive Desktop application, a functional TUI/CLI for terminal-based workflows, and a powerful Eval framework for reproducible benchmarking.
• Controlled Permissions: Implement fine-grained control over agent capabilities, ensuring they only access and operate within designated boundaries.
• Extensible Tooling: Supports a range of local tools (e.g., Read, Write, Edit, Bash, Glob, Grep) with robust schema validation, dynamic availability, and error handling.

Typical Use Cases

• Project Analysis and Refactoring: Agents can inspect codebases, identify issues, propose refactors, and even execute changes with user-approved permissions.
• Automated Content Generation: Produce reports, documentation, creative writing, or code snippets based on complex instructions and project context.
• Reproducible AI Experiments: Utilize the Eval framework to benchmark different agent configurations, models, or prompts across defined tasks and repetitions, ensuring scientific rigor.
• Complex Workflow Automation: Orchestrate multi-step processes involving tool execution, model interaction, and artifact generation, with full visibility into each step.
• Personalized Development Environment: Integrate local models and custom tools to create a highly tailored AI assistant that deeply understands and interacts with your development workflow.