Apache Maka
An open-source workspace for building and evaluating AI agents that records every action — model messages, tool calls, results — in an append-only log, so a session can be paused, resumed, or replayed exactly.
🔗 Visit Apache Maka
Description
Building an AI agent that calls tools and takes multi-step actions is hard to debug when you can't see exactly what it did, in what order, or why it failed halfway through. Apache Maka is a workspace for that: every message, tool call, and result gets written to a durable log the moment it happens, so you can pick up a crashed session exactly where it left off or trace back through a long run to see precisely what the agent tried.
It runs locally (Desktop via Electron, CLI, or TUI) with built-in tools like Read, Write, Edit, Bash, Glob, and Grep, session branching and search, and multi-model support with streaming output. Its evaluation mode lets you set up reproducible benchmarking experiments — useful for comparing how different models or prompts perform on the same task. Under Apache Software Foundation incubation (not yet fully endorsed by the ASF), built on TypeScript/Node.js with SQLite storage; currently macOS-first, with Windows in unsigned preview and Linux not yet supported.
💬 Our review
The short version: for developers building or evaluating AI agents who need to know exactly what happened during a run — not just the final output — Maka's append-only event log is a genuinely useful piece of infrastructure, and it's free.
Most agent frameworks (LangChain's agent tooling, AutoGen, CrewAI) focus on making agents easy to build; Maka's focus is narrower and more specific — auditability and reproducibility of what an agent actually did, which matters more once you're debugging production failures or running controlled evaluations than during initial prototyping. Being an early-stage Apache Incubator project is a real caveat: data formats can still change, platform support is limited to macOS today (Windows preview is unsigned, Linux isn't supported yet), and the ASF explicitly notes the project isn't fully endorsed yet. For teams that need serious agent observability today on a supported platform, it's worth adopting with eyes open; teams on Linux, or those needing a mature, stable API, should wait for it to graduate from incubation.
💰 Pricing
📊 Global score
🤖 AI-enriched data
100% gratuit, licence Apache 2.0. Projet en incubation Apache (ASF), formats de données susceptibles d'évoluer.
Pros
Journal d'événements append-only complet : auditabilité totale et reprise après crash
Fonctionne 100% en local, données conservées chez l'utilisateur
Trois interfaces : Desktop (Electron), CLI/TUI, mode évaluation
Branchement de sessions pour explorer sans perdre l'historique
Projet actif sous incubation Apache, gouvernance établie
Cons
Projet en incubation précoce : les formats de données peuvent encore changer
Support plateforme limité : macOS Apple Silicon en priorité, Windows en preview non signée, Linux pas encore supporté
Nécessite Node.js v22.19+
Pas encore pleinement validé par l'Apache Software Foundation