TL;DR
Thorsten Meyer AI has introduced Threlmark, an open-source roadmap tool built around a plain JSON file stored on the user’s disk. The project frames that file as the shared contract for humans, tools and agents, but details such as schema stability, collaboration handling and adoption are still unclear.
Thorsten Meyer AI has introduced Threlmark, an open-source roadmap tool that stores the roadmap as a plain JSON file on the user’s own disk, a design the project describes as “disk is the contract.” The announcement matters because it positions Threlmark against SaaS roadmap systems by making the file, rather than a hosted database or API, the shared point of control.
According to the source material, Threlmark’s kanban board is a view over a local roadmap.json file. The project says humans, external tools and automated agents can read from and write to the same file directly, without relying on a vendor API, rate limits, webhooks or hosted account permissions.
The tool is described as open source under the MIT license and available through threlmark.com. The source also says Threlmark is part of Thorsten Meyer AI’s 19-day Built in Public series and sits in an “operator portfolio” alongside related products, including IdeaClyst.
Threlmark is presented as a scored kanban board, meaning roadmap items carry priority scores rather than appearing as an undifferentiated list. The source says verdicts from IdeaClyst can become scored roadmap items in Threlmark, but it does not provide technical release notes, a version number or independent usage data.
Threlmark — disk is the contract
The roadmap is a plain JSON file on your disk. The board is just a view over it — and your tools and your agents read and write the same file directly.
Independent commentary, produced with AI assistance under human editorial oversight. The views are the author’s own and may change. Threlmark is open source under MIT, provided “as is” without warranty; see the repository LICENSE. Automated agents that read and write the roadmap file may introduce errors — treat agent writes as changes to review, not facts to trust. Product and company names are trademarks of their respective owners; mention does not imply endorsement.
A Roadmap Outside SaaS
The central claim is about control of operational data. If the roadmap is a plain file with a known structure, users can inspect it, version it, back it up and move it between tools without waiting for a platform export or API access. That could appeal to developers, solo operators and AI-heavy workflows where local files are easier to automate than hosted products.
The source frames the choice as an anti-lock-in move. That claim is plausible because JSON is widely supported, but the practical value depends on how stable the file format is, how well Threlmark handles invalid edits, and whether other tools actually adopt the same structure.
The agent angle is also central. Threlmark is built around the idea that AI agents can write roadmap changes into the same file humans use. The project’s own disclaimer says those writes should be reviewed, not treated as facts.
JSON file editor for project management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Built In Public Day Seven
The Threlmark post is labeled Day 7 of a 19-part Built in Public sequence from Thorsten Meyer AI. In that series, Threlmark appears to occupy the planning layer: IdeaClyst produces verdicts, and Threlmark turns selected ideas into ranked work.
The source contrasts Threlmark with typical roadmap tools that store plans in hosted databases behind proprietary APIs. Threlmark’s answer is to make the file format the integration point. In the source’s wording, “anything that reads JSON is a client.”
The announcement also says the software is provided under MIT and “as is” without warranty. That matters for readers evaluating whether this is a product ready for daily operational use or a still-developing open-source project.
Open Questions On Readiness
The source does not state whether Threlmark has a stable public schema, packaged release, installation path, multi-user sync model or conflict-resolution system. It is also unclear how the tool validates agent-written changes or recovers from malformed JSON.
No outside adoption figures, contributor counts, security review or third-party integrations are provided in the source material. Claims about durability, interoperability and lower lock-in are the project’s stated rationale, not independently confirmed performance outcomes.
Schema And Adoption Tests
The next milestones to watch are whether Threlmark publishes a clear file schema, documents safe edit patterns for agents, and shows examples of external tools reading and writing the roadmap file. If the project wants the file to function as a contract, stability and validation will matter as much as the interface.
Readers following the Built in Public series should also watch how Threlmark connects with IdeaClyst and the rest of the portfolio over the remaining posts. For now, the confirmed development is the project announcement and its file-first design; broader utility remains to be shown through releases and usage.
Key Questions
What is Threlmark?
Threlmark is described by Thorsten Meyer AI as an MIT-licensed scored kanban roadmap tool that stores its data in a local JSON file.
What does “disk is the contract” mean?
It means the roadmap file itself is the shared interface. Instead of requiring a SaaS API, tools and agents are expected to read and write the same JSON file.
Is Threlmark confirmed to be production-ready?
The source does not provide a production-readiness claim, version number or independent adoption data. It describes the project, its design thesis and its MIT license.
Why use scores on roadmap items?
The project says scoring forces ranked priorities, so roadmap items are ordered by visible trade-offs rather than all being marked high priority.
What should users be careful about?
The source warns that automated agents may introduce errors when writing to the roadmap file. Agent changes should be reviewed before being treated as reliable.
Source: Thorsten Meyer AI
