Apache Maka (Incubating) is an open-source, local-first AI agent workspace and auditable runtime developed under the Apache Software Foundation (ASF). Designed to solve enterprise compliance, security, and reproducibility challenges in autonomous software engineering, Maka provides a local-first environment where model messages, tool executions, and security permissions are recorded in a tamper-proof, append-only event log.
Unlike traditional agent platforms that execute arbitrary shell scripts directly on host environments, Apache Maka operates a sandboxed execution engine with fine-grained capability security. Developers and compliance teams can inspect every filesystem read, network socket call, and code mutation before authorization, with full cryptographic verification across all agent sessions.
A standout capability of Apache Maka is its deterministic session replay. Because all state transitions and model outputs are persisted to an append-only event log, engineers can step backward and forward through multi-hour agent reasoning cycles, fork execution trees at specific decision points, and audit exact prompt-to-code provenance across complex engineering tasks.
Licensed under the Apache-2.0 open-source license ($0 software cost), Apache Maka delivers both an interactive desktop application and a terminal TUI/CLI. It provides an enterprise-ready, self-hosted foundation for organizations seeking sovereign AI agent governance without sending telemetry or source code to third-party cloud platforms.
