📊 Full opportunity report: Disk Is the Contract: Inside Threlmark’s Local-First Architecture on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

Threlmark’s latest architecture design centers on using disk-stored JSON files as the definitive source of truth, eliminating the need for a server or database. This approach allows external tools and AI agents to interact directly with project data, making workflows more portable, safe, and open.

Threlmark is built as a Next.js app that manages project data entirely through files stored on disk, specifically JSON files in a designated directory (~/.threlmark). The core principle is that the disk layout itself functions as the API, with no server or cloud dependency. This design choice enables multiple tools to read and write data directly, fostering interoperability and avoiding vendor lock-in. The main directory contains a manifest (threlmark.json), dependency graphs (links.json), and per-project folders housing metadata, lane configurations, and individual task cards as separate files. External suggestions, agent reports, and handoffs are stored in dedicated subfolders, maintaining a clear, inspectable structure. Safety is achieved through atomic file writes—writing to temporary files before renaming to prevent corruption—and through a read-merge-write process that preserves data integrity and forward compatibility. The system’s architecture ensures restartability, as all state is stored in files, not memory, enabling seamless recovery and migration.

Why Disk-Based Data Matters for Project Management

This architecture shifts the paradigm of project management tools by prioritizing data portability, transparency, and external collaboration. It enables users to integrate their workflows with external tools and AI agents without relying on proprietary servers or lock-in, fostering a more open ecosystem. Additionally, the safety mechanisms reduce risks of data corruption, making workflows more reliable. For developers and organizations, this approach simplifies backup, migration, and interoperability, potentially influencing future design choices in productivity tools.

The Evolution of Local-First Architectures in Productivity Tools

Traditional project management tools often depend on centralized servers, databases, or cloud services, which can limit portability and control. Threlmark’s approach builds on the growing trend of local-first software, emphasizing user ownership of data and resilience. Its design echoes principles seen in other file-based systems, but applies them specifically to complex workflows involving AI agents and multi-tool integrations. The decision to treat the disk layout as the API is a deliberate departure from conventional database-driven architectures, aiming to empower users and developers alike.

“The core idea is simple: the on-disk layout is the API, which cascades into how concurrency, external participation, and AI automation are handled.”

— Thorsten Meyer

Unresolved Questions About External Tool Integration

While the architecture is designed for openness, it is still unclear how extensively external tools and third-party AI agents are currently integrated or how they will evolve in practice. The robustness of conflict resolution and security measures in collaborative scenarios remains to be tested in real-world use cases.

Next Steps for Broader Adoption and Tool Support

Threlmark plans to release detailed documentation and examples to facilitate external tool integration. Future development may include formal APIs or plugins to streamline collaboration, as well as user feedback to refine safety and concurrency mechanisms. Monitoring how the community adopts and adapts this architecture will shape its evolution.

Key Questions

How does Threlmark ensure data safety without a database?

It uses atomic file writes—writing updates to temporary files before renaming them—and a read-merge-write process that preserves data integrity and forward compatibility.

Can external tools modify project data without conflicts?

Yes, since each task card is stored as a separate JSON file, external tools can update individual items atomically without race conditions, and the system self-heals lane ordering on read.

What are the advantages of this disk-based approach?

It offers portability, transparency, interoperability, and restartability, enabling easy backups, migrations, and integrations with any tool that can read or write JSON files.

Is this architecture suitable for large or complex projects?

The design is optimized for small to medium-sized workflows where transparency and external collaboration are priorities. Scalability considerations are still under exploration.

Source: ThorstenMeyerAI.com