FAQ
Privacy and data
Does my code leave my machine?
spec-editor does not send your code to us. The tool runs on your machine, and the model is files in your repository. Text does go to the model provider you choose when a step calls a model: that is your Claude Code subscription, your API key or a local model, under that provider’s own terms. With a local model through Ollama the prompts stay on your machine. The steps in the quickstart call no model at all.
Where is the data stored?
In your project folder, as Markdown files under aspects/, next to your code and in the same git repository. The tool also keeps a journal of events and a local index inside the project. There is no database server to run. Settings are in local.yaml in the project.
Is there a hosted version?
No. spec-editor runs on your machine. A hosted view of a model, to read and share it without installing anything, is coming soon.
Models and money
Which models and subscriptions work?
Three kinds of provider: the local Claude Code CLI signed in with your own subscription (no API key), API keys through LiteLLM, and local models through Ollama. The open core does not need any of them for the model itself, the tools around it or the MCP server. Steps that call a model do.
Do you sell tokens?
No. You pay your model provider directly, with your subscription or your own keys. Pro adds cost control (a difficulty classifier, a ladder of models, budget guards, a pool of subscription accounts) so that your spend goes where it helps. See spec-editor Pro.
What is free and what is paid?
The open core is free under the Apache-2.0 license: the requirements model, the methodology, validation, links between code and requirements, export, the CLI, the MCP server, the local web interface and the editor integrations. spec-editor Pro is the paid plugin that adds the delivery team: roles, review stages, test stands, cost control, breakpoints, Reengineer, coverage and local code search. Pro is in early access; see pricing.
Using it
Can I use it on an existing codebase?
Yes. You can write the model for existing code by hand or with an external agent over MCP, and link functions to requirements with @implements. Restoring a model from existing code automatically is called Reengineer. It is part of Pro and is started from the interface.
Which languages can my project be in?
The model is Markdown and YAML, so the requirements do not depend on the language of your code. The interface of spec-editor is in English. The bundled methodology is translated into five languages. The @implements markers are plain comments, read out of the box from Python, TypeScript, JavaScript, Rust, Astro, HTML, CSS, Markdown and YAML files; you can add file types in the project settings.
Can I change the structure of the model?
Yes. The structure is data in methodology.yaml: aspects, element types, relationships. See concepts.
Comparisons
How is this different from an AGENTS.md file?
An AGENTS.md file is free text that an agent reads at the start of every session. It is useful for conventions, and it grows until nobody can say what in it is still true. spec-editor keeps requirements as separate elements with ids, statuses and acceptance criteria, and gives an agent only the elements it asks for. You can validate the model and trace code to it. The two work together: AGENTS.md can tell an agent to read the model through MCP. More in the articles.
How is this different from a spec kit?
A spec kit is a set of templates and prompts that helps an agent write a specification for one feature, as a document. spec-editor treats the specification as a structured model of the whole system that persists, is checked by tools and can be queried. A spec kit can feed spec-editor: its document is one more source. More in the articles.
Help
Where do I report a problem or ask for access?
Use the contact page.