Sign InOpen Brain
GitHubGitHub RepoNeeds Review

apache/maka

Maka gives local coding-agent runs a durable event log, sandbox boundary, recovery path, and shared runtime across desktop, CLI, and evals. It is early, source-only software with changing interfaces.

GitHub
Open Source Open MarkdownOpen JSON
Source Summary

Apache Maka records model messages, tool calls, results, permission decisions, and termination events in an **append-only local log**. Desktop, CLI, TUI, and evaluations execute through one **Runtime Host**, with sandbox approval for file writes and shell commands.

Practical Implication

Builders can treat prompts as projections of durable run history: trim old tool output from context without deleting evidence, inspect failures, recover crashes, and compare reproducible experiment cells. Its graph mode also uses isolated Git worktrees for parallel implementation.

Agent-Ready Context
Apache Maka records model messages, tool calls, results, permission decisions, and termination events in an **append-only local log**. Desktop, CLI, TUI, and evaluations execute through one **Runtime Host**, with sandbox approval for file writes and shell commands.

Builders can treat prompts as projections of durable run history: trim old tool output from context without deleting evidence, inspect failures, recover crashes, and compare reproducible experiment cells. Its graph mode also uses isolated Git worktrees for parallel implementation.

This is an **Apache Incubator** project with no approved Apache release. The desktop build currently supports **Apple Silicon macOS**; formats and commands may change, Windows is an unsigned preview, and secrets remain plaintext to the local OS account.
Connected Context · Feed7 Judgment

Maka turns several proposed reliability primitives into one local runtime: durable event history, permissioned execution, replayable evaluations, crash recovery, and isolated parallel work. It strengthens the case for event-sourced harnesses while narrowing immediate adoption through incubator status, changing interfaces, limited platform support, and plaintext local secrets.

Agent Frameworks Considered Harmful — Rémi Louf, .txtMaka directly implements the candidate’s durable-event and replay architecture through an append-only log shared across interfaces and evaluations.Multiplayer agentic engineering — Arjun Singh, SuperconductorIts Runtime Host, sandbox approvals, visible history, and isolated Git worktrees provide a local implementation of the candidate’s inspectable, isolated parallel-agent workflow.FinOps for AI Agents: Who Spent All the Tokens? — Tisha Chawla & Susheem Koul, MicrosoftKeeping evidence while trimming old tool output separates audit history from active context, supporting run-level context control without deleting the data needed for cost and failure analysis.Building a software factory for AI SDKMaka supplies runtime primitives for the factory pattern—isolated worktrees, recorded evidence, and permission gates—while leaving human review and merge authority outside the runtime.
Context Map
agentcoding#harness-engineering#agent-reliability#sandboxing
Uncertainty
This is an **Apache Incubator** project with no approved Apache release. The desktop build currently supports **Apple Silicon macOS**; formats and commands may change, Windows is an unsigned preview, and secrets remain plaintext to the local OS account.