Apache Maka (Incubating) is an open-source agent workspace that keeps a complete, replayable record of everything an AI agent does. Developed under the Apache Software Foundation’s incubator programme, it gives developers a high-performance desktop environment for running tool-using agents locally, with event sourcing at its core: every action, decision and tool call is captured as an event that can be inspected and audited afterwards.
Most agent tools leave you with a chat transcript and little else. Maka takes a different approach, persisting the full event history of every session so runs can be replayed step by step, compared against each other and debugged when something goes wrong. It combines a local-first Electron desktop app with a command-line interface, both written in TypeScript, and is designed for teams that need transparency and accountability from autonomous agents rather than a black box. The project has earned more than 5,800 stars on GitHub and was still under active development in 2026. Because the record is stored as an append-only event log rather than a summary, teams can also export it for compliance reviews or post-mortems, and new sessions can build on the verified history of earlier ones instead of starting from a blank slate.
Free. Apache Maka is open source under the Apache-2.0 license and runs locally on your own machine, with no paid tiers or hosted service required.
Developers and engineering teams who run local AI agents and need a trustworthy audit trail of what those agents did, including regulated environments where agent behaviour must be reviewable after the fact.