TL;DR
Threlmark’s published architecture notes describe a local-first project management system built on plain JSON files rather than a traditional database. The report says the on-disk layout acts as the app’s API, but release status, adoption, and real-world performance remain unclear.
Threlmark’s latest technical write-up describes a project management app that treats local disk storage as the system of record, using plain JSON files instead of a database, cloud backend or user accounts. The design matters because it would make project data inspectable, portable and available offline if implemented as described.
The source material identifies Threlmark as a Next.js and TypeScript app built around JSON files stored on disk. It says the default data root is ~/.threlmark, with files for the manifest, dependency graph, projects, boards, items, suggestions, handoffs, reports and archived data.
The central confirmed design claim in the material is that the on-disk layout acts as the application contract. Individual cards are stored as separate item files, while board ordering is kept in board.json. According to the write-up, external tools can add or edit item files directly, and Threlmark reconciles board state when it reads the files.
The article also says the system uses atomic writes: data is written to a temporary file in the same directory, flushed, then renamed over the target file. The source describes this as a guard against partial writes, with the aim of leaving either the previous full file or the new full file after an interruption.
Why It Matters
The architecture is relevant for developers and technical teams evaluating local-first software, especially tools that need to work with AI agents, command-line workflows or version control. A file-based contract can make data easier to inspect with standard tools, move between systems and back up without a vendor-specific export path.
The approach also changes how integrations work. Rather than requiring every outside tool to use a proprietary API, the source says any tool that can read and write the expected files can participate. That could lower integration friction, although reliability depends on how well the file contract, conflict handling and validation are maintained.
plain JSON file project management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
The write-up is presented as part two of a Threlmark series and frames the product as a local-first alternative to server-centered project management tools. It describes a structure in which each project has metadata, board state, one file per card, an inbox-style suggestions folder, agent handoff records and report drop zones.
The material places special weight on AI agent workflows. It says a card can be handed off, completed by an agent, reported back through either a REST endpoint or a report file, and then moved based on that report. The report path is described as a fallback for cases where the server is unavailable when the agent finishes.
Several values are described as derived rather than stored. Priority, stale age, cycle time, throughput and work-in-progress counts are computed from item state or event history, according to the source. That design is meant to reduce drift between displayed metrics and the underlying files.
“The on-disk layout IS the API.”
— Thorsten Meyer AI source material
“There is no server-of-record — the files are the record.”
— Thorsten Meyer AI source material
“Every artifact is a file you can cat, diff, grep, commit.”
— Thorsten Meyer AI source material
“Anything computable from item state is computed.”
— Thorsten Meyer AI source material
Real-World Android App Projects with Kotlin and Jetpack Compose: Build Production-Style Android Apps with Modern Architecture, API Integration, State Management, Local Data Storage, Practical Projects
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
The source material does not specify Threlmark’s current public release status, user adoption, production deployments or benchmark results. It is also not clear from the provided material how the system handles cross-device conflicts, simultaneous edits from sync tools, malformed external writes, permission errors or security boundaries for agent reports.
offline project tracking tool
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next points to watch are whether Threlmark publishes implementation details, tests, release notes or user-facing builds that show how the file contract behaves under real workloads. Developers evaluating the approach will need to see conflict behavior, validation rules and agent report handling before judging whether the architecture is reliable beyond the design description.
AmpOhm360 Waveshare USB 5V 120mm Quiet Cooling Fan, 3-Speed Adjustable PWM Control, High Airflow Anti-Vibration Design, Compatible with Development Board and Microcontroller Projects, SKU-34408
【3-Speed PWM Control】Easily switch between high, medium, and low speeds for optimal cooling performance, with one-touch shutdown for…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news development?
Thorsten Meyer AI published a technical report explaining Threlmark’s local-first architecture, with local JSON files acting as the primary data record.
Does Threlmark use a database?
According to the source material, Threlmark does not use a traditional database, cloud backend or account system. It stores project data as plain files under a local data root.
Why does one file per card matter?
The source says storing each card in its own file reduces write collisions and lets external tools create or modify cards without directly changing board ordering.
How do AI agents fit into the system?
The write-up says agents can receive handoffs and report completion through REST or by dropping a report file. A done report can move a card forward in the workflow.
What is still unknown?
The provided material does not confirm release timing, adoption, performance, security review, or how the system handles complex sync conflicts across machines.
Source: Thorsten Meyer AI
