# Vondet

> However much you talk, the plan stays one. A board your AI writes to. Instead of a fresh summary every conversation, it stacks into one plan.

## What it is

- You talk to your AI: In the app you already use. Your AI reads the board as it goes.
- Your AI writes here: Goals, moves, reasoning, and what would reverse them.
- Open it only when you want to: You don't have to come here — the conversation picks up from the plan. Open it to see what's decided and how far it got. To change anything, just tell your AI.

## What is not here

There is no box for writing a plan from scratch. No chat panel, no ticket form. You correct; the AI does the writing.

## When the plan stays one, three things go your way.

- **Away a while? Pick up where it was.** — However long you're away, the plan is one. The next AI reads the same board and just continues.
- **Overturn the AI's call, reasons and all.** — Every call carries its reasoning and what would reverse it. Disagree, and a word rewrites it. The earlier call stays on record.
- **No two sessions collide.** — A later you — or another session's AI — reads the same single plan. No colliding behind each other's backs.

## Status

- Sign-ups are open. You can start from the sign-in page.
- Export: Yes. From account settings you can export the whole board as JSON in one click, on the free plan too. The contents are yours and we do not intend to lock them in.

## Common questions

### Does it work with anything other than Claude?

Claude is the only one we have verified so far. The connection uses MCP, an open method, so other assistants may well work — we have not checked.

### Does it read my code?

No. It never touches anything on your machine. What is on the board is what the AI wrote there, plus whatever you change yourself — titles, reasoning, conditions, and links you allowed.

### Can I take my data out?

Yes. From account settings you can export the whole board as JSON in one click, on the free plan too. The contents are yours and we do not intend to lock them in.

## Links

- [Landing page](https://vondet.com/en)
- [Pricing](https://vondet.com/en/pricing)
- [Writing](https://vondet.com/en/blog)
- [Claude Code todo list: the day I stopped rebuilding it from memory](https://vondet.com/en/blog/claude-code-todo-list) — A builder's story of a Claude Code todo list that grew past memory: how much was done, what was left, what contradicted what, and what a board changed.
- [Run a small AI coding productivity study, then audit your own numbers](https://vondet.com/en/blog/ai-coding-productivity-study) — Five practices for measuring what a coding agent does in your own repo, plus an audit built from eight claims our own articles could not defend.
- [Spec-driven development with Claude Code: check that the spec arrived](https://vondet.com/en/blog/spec-driven-development-with-claude-code) — Five practices that make spec-driven development work with coding agents, and the one gap none of them close: the copy your agent runs can be days old.
- [Running Claude Code agents in parallel: what git stops, and what it doesn't](https://vondet.com/en/blog/running-claude-code-agents-in-parallel) — git worktree, split roles, a file boundary, visible windows: four things that make parallel Claude Code agents work — and the one problem none of them fix.
- [Building a Claude Code handoff, and measuring what it carries](https://vondet.com/en/blog/claude-code-handoff) — A Claude Code handoff already works well enough that people ship it as a handoff.md, a slash command, a skill. Paste nothing, and half of it already arrives anyway.
- [Why Claude Code ignores the rules you wrote for it](https://vondet.com/en/blog/why-your-agent-ignores-your-rules) — We added one line to a CLAUDE.md and a rule went from 0 in 7 runs to 2 in 2. What changed, along with where the rule lived, was whether Claude Code could check the condition while it was writing.