What is Legevio?

Legevio keeps a project’s intent, order and decisions in one place: the tasks on its board, the results waiting for review, and the Project Memory your team has accepted. The AI you already use connects to it, reads what a task needs, and hands its work back for you to review.

Two things change when you work this way.

• Nothing gets lost between chats. A new chat or agent starts from the project’s state page and the decisions you accepted, not from a pile of handoff notes.
• You run the project without reading the code. You see what is done and what is waiting for you, review each result with its evidence, and decide what becomes the project’s truth.

Tasks

Ready to hand off

Store page capsule art brief#29Creator

Design Mira's turnaround#5Creator

In Progress

Prototype the tide simulation#26Game Designer

Review

Color-script Gullwatch at dusk#3Creator

Level 3 puzzle: the Tide Clerk's ledger#23Game Designer

InboxMessagesDraftsNotes

Color-script Gullwatch at duskIn Project Memory

Outside AI via MCP · Claude Code#3Color-script Gullwatch at dusk

Git references — reported by the executor, not verified by Legevio

Commitc41e7a2

Tests12 pass

Files into Context · Style guideChange

In Project Memory· Context · Style guide

#29Store page capsule art briefCreator

Task ContextResultDiscussion

Included in this brief4

• Lanternfall on one pageOverview
• Palette: dusk teal and lamp amberContext · Style guide
• Decision: no combat — every encounter is a puzzleDecisions
• Color-script Gullwatch at duskContext · Style guide

Sample data

When it is worth it

For any project longer than a couple of days, set up from day one. For a one-evening task, it is not worth it. Legevio keeps intent, order and decisions. It does not replace your repository, your code or your tests.

What Legevio is not

• Not a model. Your AI keeps running where it runs today and connects to Legevio over MCP.
• Not a code host. Commits and pull requests stay in your repository. Your AI reports the ones it made with each result, so you review the work with its code references.
• Not automatic truth. Nothing an AI writes becomes Project Memory on its own. A person accepts it, or an acceptance rule the project owner turned on does, and either can be undone.

Start here

1. Connect your AI. Add Legevio to the AI you already use, so it can read your project and hand its work back for review.
2. Connect a computer. Chat with Claude Code or Codex on your own computer from Legevio, with your own sign-in and subscription.
3. Run your first project. From an empty workspace to a first accepted decision that the next task receives.

Then read how the pieces fit: Project Memory and acceptance, What the AI receives, Roles and the team and the Inbox.

Five things to set up on day one

Most early problems come from starting late: decisions stay in old chats, and early commits can no longer be linked to their tasks.

1. A project state page. Every chat and agent starts from it.
2. Review, then accept. Results are reviewed before they count. Turn on an acceptance rule only where you want one.
3. The task number in every commit, and the commits reported with each hand-in, so every result shows the code it belongs to.
4. Every decision from chat becomes a Project Memory record, so the next chat sees it.
5. Handoff files for procedures only, such as how to build or how to test. Status lives in Legevio.