TL;DR
Thorsten Meyer AI has described Threlmark as a local-first project management app built on plain JSON files rather than a traditional database. The report says the on-disk layout is treated as the system contract, with atomic writes, one file per card and file-based AI-agent reports. Adoption, performance data and release timing remain unclear from the provided material.
Thorsten Meyer AI has published a technical report describing Threlmark, a Next.js and TypeScript project management app that uses plain JSON files on local disk as its system of record instead of a database, cloud service or user accounts, a design choice that affects how the tool stores cards, syncs with other software and supports AI-agent handoffs.
According to the report, Threlmark’s default data root is ~/.threlmark. The directory contains a manifest, a links graph, project folders, shared items and an archive. Within each project, metadata, lane ordering, individual item files, suggestions, handoffs, reports and a human-readable roadmap are stored as files. The source describes that layout as the API used by both the interface and outside tools.
The report says Threlmark relies on two safety patterns rather than a database. Writes are made to a temporary file in the same directory, followed by fsync and rename over the target file. The source states that rename is atomic on one filesystem, meaning a crash should leave either the old file or the new one rather than a partial write. The second pattern is one JSON file per card, while lane order lives in board.json and is reconciled on read.
The source also says values that can be calculated are not stored separately. Priority, stale age, cycle time, throughput and work-in-progress counts are derived from item state and append-only history. For AI-agent work, the report describes a handoff and report loop: an agent can report completion through a REST endpoint or by dropping a report file into a project folder, and a done report can move a card to Done.
Why It Matters
The design matters because it makes project data inspectable and portable in ways that cloud-first or database-backed tools often do not. Files can be read, diffed, backed up, committed to version control or moved between systems with ordinary developer tools. That could appeal to users who want offline access, low lock-in and direct control over their working data.
The approach also changes integration costs. If the file layout is stable, another tool does not need a private database driver or vendor API to participate; it can read and write the same files under the same rules. The tradeoff is that the file contract has to carry much of the discipline normally handled by a database, including write safety, reconciliation and compatibility as the app evolves.

Information Technology Project Management (MindTap Course List)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Threlmark is described in the source as an MIT-licensed Next.js app built on TypeScript and JSON-on-disk storage. The report is labeled part two of the Threlmark series and focuses on architecture rather than a product launch, customer rollout or funding event.
The report contrasts the per-card file model with a single roadmap JSON array, which it says can create write races when multiple tools modify state. In Threlmark’s design, external tools are expected to write item files while the board repairs ordering and state during reads. The source also says unknown keys are preserved, which is meant to let future tools add data without breaking older readers.
“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
“Just use files is easy to get wrong.”
— Thorsten Meyer AI source material
“Unknown keys are preserved.”
— Thorsten Meyer AI source material

Forvencer 12 Pocket Project Organizer, 1/6-cut Tab Binder Organizer with Sticky Labels, Multi Pocket Folder with Zipper Pouch, Folder Binder Spiral Pocket Notebook, Office Supplies, Letter Size, Blue
High-efficiency Paper Organizer: 12 pockets, 6 dividers with 1/6-cut assorted tabs in bright colors, and 36 blank viewable…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
The source does not state whether Threlmark is already in broad production use, how many external tools have implemented the file contract, or how conflicts are handled when sync tools edit the same item file on different machines. It also does not provide performance measurements, security review findings or a dated release plan. Those points remain unverified from the provided material.
local-first task management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next milestone is likely more implementation detail, tests or release notes showing how the file contract behaves across real projects, sync services and AI-agent workflows. No specific date was provided in the source material.

Planning and Control Using Oracle Primavera P6 Versions 18 to 23 PPM Professional
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news here?
The news is that Thorsten Meyer AI has published a technical report describing Threlmark’s local-first architecture, centered on local JSON files as the app’s authoritative data layer.
Does Threlmark use a database?
According to the source, Threlmark does not rely on a traditional database, cloud server or user accounts for its core record. It stores project data in files under ~/.threlmark.
How does the design protect files from partial writes?
The report says Threlmark writes to a temporary file, fsyncs it and then renames it over the target. The source presents that pattern as a way to avoid half-written JSON files after a crash.
Can other tools work with Threlmark data?
The source says yes, if they follow the file contract. External tools can read and write the same file layout, including item files and report drop zones.
What happens next?
The open question is whether the architecture is shown in wider use, with release notes, benchmarks, conflict-handling details and examples of outside tools using the file contract.
Source: Thorsten Meyer AI