maka vs sandbase-harness
Local AI agent runtimes compared
maka and sandbase-harness are TypeScript projects for running AI agents on local machines or self-managed infrastructure. maka centers on a shared runtime with an append-only activity log and desktop, TUI, and CLI clients; sandbase-harness focuses on operating persistent sessions with sandbox, memory, credentials, and MCP support.
maka: Keep a Complete Record of AI Agent Work
Apache Maka is a high-performance agent workspace that records every model message, tool call, permission decision, and turn ending. It runs as a desktop app, with TUI and CLI clients, on macOS and preview builds for Windows and Linux.
sandbase-harness: Run Self-Hosted AI Agents with Sandboxed Tools
SandBase Harness is a local-first AI agent runtime and MCP bridge, with a built-in Console and Anthropic Claude Managed Agents-style API. It provides sessions, sandboxed tools, memory, credential controls, and audit and replay on your own machine or infrastructure.
| maka | sandbase-harness | |
|---|---|---|
| Language | TypeScript | TypeScript |
| License | Apache-2.0 | Apache-2.0 |
| Stars | 5.8k | 639 |
| Forks | 557 | 80 |
| Last analyzed | Oct 19, 2026 | Oct 19, 2026 |
Key differences
- maka records model messages, tool calls, and permission decisions as append-only RuntimeEvents; sandbase-harness offers persistent sessions with event history, audit, and replay.
- maka provides a desktop app, TUI, and CLI backed by one Runtime Host; sandbase-harness offers a Console, CLI, TypeScript SDK, and a `/v1` API.
- maka supports user-provided cloud APIs, local models, and compatible gateways; sandbase-harness lists OpenAI, Anthropic, MiniMax, and OpenAI-compatible providers, and requires a provider API key.
- maka keeps session data and run records locally; sandbase-harness uses SQLite metadata and local file storage by default, and also supports self-managed deployments with Docker, Kubernetes, or workers.
- maka is marked as incubating, with Windows and Linux support described as previews; sandbase-harness is less than six months old and has documented compatibility limitations and backend-dependent sandbox boundaries.
- Both use TypeScript and Apache-2.0; maka has 5.8k stars and 551 forks, while sandbase-harness has 707 stars and 78 forks.
Choose maka if you…
- want a desktop agent workspace with matching TUI and CLI clients.
- need append-only records of agent messages, tool calls, and permission decisions.
- want to connect a cloud API, local model, or compatible gateway and inspect evaluation results.
Choose sandbase-harness if you…
- need persistent sessions with replay, memory, credential vaults, and MCP toolsets.
- plan to run agents on self-managed infrastructure using Docker, Kubernetes, or workers.
- want a Console, TypeScript SDK, and `/v1` API for operating agent sessions.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.