Welcome

Jumpspace is repo-local intent memory for AI coding agents.

It turns source-controlled Markdown into durable decisions, rationale, rejected alternatives, and scoped path checks that agents can read before editing code.

The First Win

The first win is not a perfect knowledge graph. The first win is giving an agent a grounded way to answer:

• Which decisions apply to the files I am about to edit?
• Why did the team choose this direction?
• Which tempting alternatives were already rejected?
• Should this branch add one durable intent, or is the code enough?
• What can the next agent learn without relying on chat history?

Who It Is For

Use Jumpspace when:

• your team uses AI coding agents across sessions or branches
• product or architecture intent is easy to lose between sessions
• issue trackers and commits do not explain why decisions were made
• agents keep rediscovering rejected alternatives
• you want repo-local state instead of hosted memory

Skip it for tiny throwaway repos where a README and grep are enough.

What To Run First

For a new or clean repo:

npm install -D @jumpspace/cli
npx @jumpspace/cli init --auto
npx @jumpspace/cli add-skill --all
npx @jumpspace/cli intent list
npx @jumpspace/cli intent validate --json

Then add one intent under documentation/intents/ for a decision code cannot explain, and run:

npx @jumpspace/cli intent check --for src/app/page.tsx

For teams that also want source-backed task state, continue with the advanced bootstrap workflow.

For agent setup, go to Agent Setup.