TARDIS v0.3.6: Local-First Multi-Agent AI Office

What changed

The latest release, v0.3.6, of the TARDIS project brings a fresh update to its local-first, self-hosted multi-agent office. This project, written in TypeScript, is designed to integrate with a range of AI models, including Claude Code, Codex, Grok, GLM, and other local models. The repository indicates a recent push and release on September 28, 2026, suggesting active development. TARDIS aims to provide a unified environment for managing multiple AI agents.

Why it matters for builders

For developers, TARDIS offers a self-hosted solution for building and managing complex AI agent systems. The local-first architecture means that core functionalities operate on the user's machine, potentially reducing latency and enhancing data privacy. Support for diverse AI models allows for flexibility in choosing the best tools for specific tasks within the multi-agent setup. The project's use of TypeScript and React suggests a modern development stack, potentially making it easier for web developers to contribute or extend.

Practical impact

Builders interested in experimenting with multi-agent AI systems can explore TARDIS for a self-contained environment. The self-hosted aspect is particularly relevant for projects requiring strict data governance or offline capabilities. Developers can investigate the project's structure to understand how different AI models are integrated and orchestrated. Given the recent release, it presents an opportunity to engage with a project that is actively being maintained and potentially shape its future direction.

Caveats and source limits

The repository metadata is limited in detail regarding specific new features in v0.3.6, the exact architecture of the multi-agent system, or performance benchmarks. Information on setup requirements beyond the language (TypeScript) and the presence of a package.json is not provided. The project currently has a low star count (1 star) and fork count (0 forks), indicating it is in its early stages of adoption and community engagement. There are no explicit mentions of documentation or examples within the package signals. The readme_summary mentions "4 AI signals" and "5 developer signals" but does not elaborate on what these entail.