prjct/
Tell AI what to build. Know it built the right thing.
Start new thread

prjct/cloud - Change machines or agents. Keep your project's memory.

by•
prjct/cloud backs up and syncs the project context prjct/cli creates — decisions, guardrails and active work — across machines, editors and models. New laptop or a different agent? Pick up without explaining the project again. 7-day trial, no card. From US$12/mo.

Add a comment

Replies

Best
Maker
📌
We first launched prjct/cli here over 10 months ago — free, local project memory for AI coding agents. It's a very different, much bigger product now: we ship almost daily and it's built with itself. The one request that never stopped coming in all that time: "I switch laptops and agents constantly — how do I keep the project's memory with me?" That's prjct/cloud. The decisions, guardrails and active work that prjct/cli captures locally get backed up and synced — so a new machine, a fresh clone, or a different agent (Claude, Codex, Cursor, Gemini, Grok, OpenCode) picks up with the project's context already there. No re-explaining, no rebuilding from old chats. And it never needs your source code to do it — only the project knowledge you chose to capture. Your repo stays with you. Local stays free forever (MIT). Cloud is US$12/mo, and for launch I'm keeping the Founding Lifetime at US$249 one-time while a few spots are open. 7-day trial, no card.

Love the "harness is owned" framing, really resonates after watching too many sessions lose context mid-task. One concrete ask: add a dry-run mode that previews what files the agent plans to touch and what commands it intends to run before actually executing, so I can sanity-check the intent brief against reality before letting it rip on a real codebase.

The SQLite-inside approach is such a nice touch, keeps everything portable and inspectable. Been running it alongside Claude Code and the bounded RAG actually pulls relevant context instead of dumping entire files into the prompt.