Claude Code
Yesterday's plan, right where you left it.
Save your findings during a session. Tomorrow morning, open the same folder and Claude already knows what you wrapped up — no rehydration prompts, no copy-paste from old chats.
Your AI agents shouldn't start from scratch every time. BitPub gives you one private place to store your chats, skills, and context — so every agent, in every project, picks up where the last one left off. When you're ready, share it with your whole team so everyone's agents work from the same place.
Install the BitPub CLI — it's the open-source `@bitpub/cli` package on npm (source: github.com/tollbit/bitpub). Run `npm install -g @bitpub/cli && bitpub setup`, then `bitpub list` to confirm it works.
npm install -g @bitpub/cli && bitpub setup
curl -fsSL https://bitpub.io/install.sh | bash
macOS and Linux. On Windows, run inside WSL. Requires Node.js. Both paths install the same thing — @bitpub/cli from npm — and then provision a private encrypted identity. The script (readable at /install.sh) just wraps the npm install + setup.
Insights get lost in long chat threads. Useful prompts and context stay stuck on one person's laptop. The next time you use an AI, it starts completely cold and has to relearn everything. BitPub saves the context that matters, so your AI's knowledge grows over time — across sessions, across different tools, and across your entire team.
BitPub acts like a fast, offline notebook on your computer. Your AI can read from it instantly without waiting on the internet. It only connects to the cloud when you want to pull updates from your team or share your own work.
Run one simple command. You get a secure, private memory bank tied to your current folder. It works offline instantly, and it's the exact same setup whether you're working solo or with a team.
Tell your agent to save its current plan or research. If you come back to this folder weeks later, your agent can instantly recall exactly what it was doing and pick up right where it left off.
Link your workspace to your team to access shared knowledge. You can pull a team document, have your agent work on it privately, and only share the polished version back. Your private work stays private.
Set your agents to watch for changes. When a teammate uploads a new file, your agent can automatically wake up, process it, and save the results. It's the easiest way to connect multiple AIs together.
Every document, decision, chat transcript, or AI prompt gets its own stable link. Your agents always know exactly where to find the right information, and every change is tracked and versioned.
Your team probably uses different AI tools — one person loves Cursor, another uses Claude, a third prefers Gemini. BitPub gives all of them the same shared memory. Switch tools tomorrow, and yesterday's context is still right there.
Yesterday's plan, right where you left it.
Save your findings during a session. Tomorrow morning, open the same folder and Claude already knows what you wrapped up — no rehydration prompts, no copy-paste from old chats.
Composer opens with your team's memory loaded.
New chat, new file, new repo — Cursor's agent already has your team's working notes, voice guide, and shared tools. The skill installs once; everyone benefits next session.
A second pair of eyes that already knows the project.
Switch over to Gemini for a fresh take and it walks in pre-briefed — same workspace, same notes, same team context. Compare answers across models without re-explaining yourself each time.
Run from any folder. Memory comes with it.
Codex picks up the project's working notes and your team's playbook automatically — no setup per repo, no env vars to remember. Save, switch laptops, keep going.
Everything below runs on the same six verbs — save, load, list, find, sync, delete — plus the occasional flag. No orchestrator, no queue service, nothing extra to stand up.
An agent saves what it learned; days later, one list --sync rehydrates the whole folder. Never start cold.
The moment one agent updates a shared address, every other agent reads the new truth — no stale docs, no duplicated work.
Local workers watch a namespace and race to claim each job — first writer wins, the rest get a 409. No queue service to run.
An agent saves a page, a manifest, and a script at one address — and the whole team has a working tool on their next sync. No deploy, no app store.
# Monday — the agent wraps up and saves where it landed $ bitpub save auth-spec "JWT in httpOnly cookie; refresh rotation; 15m access TTL" ✓ saved → bitpub://private:you/Projects/api/auth-spec (v1) # Thursday — new chat, zero memory of Monday. one command brings it back. $ bitpub list --sync ↺ api · 3 slices auth-spec · open-questions · load-test-notes $ bitpub load auth-spec JWT in httpOnly cookie; refresh rotation; 15m access TTL
# 9:02 — cursor (Priya's agent) revises the shared sprint truth $ bitpub save bitpub://group:co.com/WorkingMemory/Sprint-14 \ "auth shipped ✓ — rate-limit spec is the blocker, needs product sign-off" ✓ saved (v18) # 9:02 — claude-code (Sam's agent) is watching, reacts on fresh truth $ bitpub sync --watch "bitpub://group:co.com/WorkingMemory/**" ● v18 Sprint-14 · cursor, just now # drops the finished auth task — starts the rate-limit spec instead
# three local workers watch the same queue worker-1 worker-2 worker-3 ↻ watching Jobs/Pending/** # a job lands — all three attempt the claim, only one wins $ bitpub save bitpub://group:co.com/Jobs/InProgress/build-417 \ "by: worker-2" --expect-version 0 worker-2 ✓ claimed (v1) worker-1 ✗ 409 — taken worker-3 ✗ 409 — taken
Pulls your notes, the latest email thread, and open action items into a one-page brief — right before the call.
Tell your AI agent to install BitPub. One sentence. Your agents start remembering. Your team starts sharing.
Install the BitPub CLI — it's the open-source `@bitpub/cli` package on npm (source: github.com/tollbit/bitpub). Run `npm install -g @bitpub/cli && bitpub setup`, then `bitpub list` to confirm it works.
npm install -g @bitpub/cli && bitpub setup
curl -fsSL https://bitpub.io/install.sh | bash
macOS and Linux. On Windows, run inside WSL. Requires Node.js. Both paths install the same thing — @bitpub/cli from npm — and then provision a private encrypted identity. The script (readable at /install.sh) just wraps the npm install + setup.