TL;DR
Threlmark has published technical details of a local-first project management architecture built on plain JSON files rather than a database, cloud service, or user accounts. The report says the on-disk file layout acts as the API, allowing the app, outside tools, and AI agents to read and write the same project state.
Threlmark has detailed a local-first architecture for its project management tool that runs as a Next.js app over plain JSON files on a user’s disk, a design the project says allows task boards, outside tools, and AI agents to share state without a database, cloud backend, or user accounts.
The source material describes Threlmark’s central design choice as treating the on-disk layout as the API. The data root defaults to ~/.threlmark, with files and folders for a manifest, dependency graph, projects, item records, suggestions, handoffs, reports, a human-readable roadmap, shared items, and archives.
According to the Threlmark technical write-up, each work item is stored as its own JSON file under a project’s item directory. Lane ordering is kept separately in board.json, while the board reconciles item files when the app reads from disk. The report says this allows an outside tool to add a card by writing an item file without editing board order directly.
The article also says Threlmark uses atomic writes: writing to a temporary file in the same directory, syncing it, then renaming it over the target file. The stated goal is to avoid partially written JSON if a process crashes during a write. The source says unknown keys are preserved, which is intended to make the file contract more tolerant of future fields and outside tools.
Why It Matters
The architecture matters because it places project data under the user’s direct control. If the design works as described, users can inspect, diff, back up, sync, or move their work data with standard file tools, rather than relying on a hosted service as the only system of record.
For developers and AI-assisted coding workflows, the larger claim is interoperability. The source says any tool in any language can join the workflow by reading and writing the same files. That could make Threlmark useful in environments where agents, scripts, editors, and project dashboards need to coordinate without API keys, account setup, or network access.
The design also addresses a common weakness in task tools used with AI agents: the handoff can happen in one place while the result is tracked somewhere else. Threlmark’s report says an agent can receive a work brief, report completion through REST or a filesystem drop zone, and move the card to Done when a completed report is ingested.
Python in Action: 60 Mini Projects to Automate Everything (Part 1): Practical CLI Tools, File Automation, and Data Cleaning with CSV, Excel, and JSON
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Threlmark is presented in the source material as a project tool focused less on card placement and more on deciding what should be done next across all active work. The article frames the tool around portfolio ranking, lane state, handoffs, reports, and derived metrics rather than a conventional hosted project database.
The report says priority is calculated on read rather than stored, using impact, evidence, fit, and effort. Other values, including work-item age, cycle time, throughput, and work-in-progress counts, are also described as derived from item state and recorded state changes. That approach is meant to reduce the chance that displayed metrics drift away from the underlying files.
The source also describes a portfolio score that ranks globally addressable items across projects with status weights. In that model, development work receives more weight than ranked ideas, while completed work receives much less, and blockers receive an added boost based on blocked counts.
“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
“A handoff is a first-class flow event”
— Thorsten Meyer AI source material
The Phoenix Project: A Novel about IT, DevOps, and Helping Your Business Win
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
The source material does not state a publication date, release version, number of users, production adoption, or independent testing results. It is also not clear from the provided material how Threlmark handles multi-device conflicts when file sync tools modify the same project data, how permissions are managed on shared machines, or how well the model performs with very large boards.
The article describes intended safety patterns, including atomic writes and one file per item, but the provided source does not include benchmark data, failure-rate data, or third-party review of the implementation.
MaxBP Tracking Discs – 10-Pack, Black
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next milestone is whether Threlmark’s file-contract model is validated through broader developer use, integrations, and agent workflows. Readers should watch for implementation details on conflict handling, sync behavior, deployment options, and examples of external tools writing to the same project files.
Mastering PowerShell Scripting: Automate repetitive tasks and simplify complex administrative tasks using PowerShell
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news development?
Threlmark has published technical details of its local-first architecture, describing a project management tool built on JSON files stored on the user’s disk rather than a hosted database.
Is Threlmark a cloud service?
According to the source material, Threlmark is designed as a Next.js app over local files, with no database, no cloud backend, and no user accounts in the described architecture.
How do outside tools interact with Threlmark?
The report says outside tools can participate by reading and writing files in the same on-disk structure. For example, a tool can add an item file, and Threlmark reconciles board state when it reads the project.
How are AI agents part of the workflow?
The source says Threlmark supports handoffs and reports. An agent can receive a brief, report completion through a REST endpoint or a report file, and a done report can move the card to Done.
What remains unknown?
The provided material does not confirm release timing, user adoption, conflict behavior under sync tools, performance at scale, or independent verification of the architecture’s reliability.
Source: Thorsten Meyer AI
