- Aron Prins just shipped v2 of the community Paperclip docs — expanded guides, more screenshots, and better adapter coverage for the open-source orchestrator that lets you run a company with AI agents.
TL;DR
Aron Prins just dropped Paperclip Docs v2 at aronprins.github.io/paperclip-docs. It's the second pass of community-maintained documentation for Paperclip — the open-source "control plane for AI-run companies" with 58.3k GitHub stars. v2 adds more screenshots, deeper guides, and improved adapter docs, one week after the v1 first pass landed on Apr 17.

What's new in v2
The announcement from Prins on X is short and direct: v2 is "expanded, improved, and better than v1", with more screenshots, detailed in-depth guides, and "many more additions and improvements to be added soon."
- More screenshots throughout the guides — v1 was text-heavy.
- Detailed, in-depth guides replacing the first-pass tutorials.
- Improved adapter documentation for plugging in different agent runtimes.
- Active roadmap — more additions and improvements on the way.
Prins has also hinted that, thanks to Paperclip creator @dotta, the community docs may become the official docs "soon."
Why it matters
Paperclip is one of the more ambitious open-source AI projects of 2026 — not an agent framework, not a workflow builder, but an entire control plane for teams of autonomous agents. The mental model is explicit: "a company you are running, not a tool you are using."
That framing is powerful but steep. You hire agents, assign roles and budgets, set goals, approve risky decisions, and audit the work — all concepts with no good single-agent analogue. A project like that lives or dies on its onboarding surface. v2 is the clearest signal yet that Paperclip's docs are catching up with its ambition.
What is Paperclip, exactly?
If you haven't seen it yet: Paperclip is an MIT-licensed Node.js server with a React UI that orchestrates teams of AI agents to run a business autonomously under human oversight. The product positioning: "the human control plane for AI labor."
Paperclip works with any agent runtime that can receive scheduled heartbeats — OpenClaw, Claude Code, Codex, Cursor, Bash CLI agents, and HTTP/webhook bots are all supported out of the box, with a plugin system for custom adapters. Core primitives worth knowing before you open the docs:
- Agents — hired bots with a role, title, monthly budget, and reporting line.
- Tasks / Issues — work items with atomic checkout so two agents never double-work.
- Heartbeats — scheduled wake-ups where an agent picks up assignments and acts.
- Budgets — hard per-agent token/cost caps; execution pauses when you hit them.
- Governance — board-level approval workflows and execution policies with audit logs.
- Workspaces — isolated git worktrees or directories where agents actually do the work.
Technical facts
| Property | Value |
|---|---|
| Repo | paperclipai/paperclip |
| License | MIT |
| Stars / forks | 58.3k / 10.1k |
| Codebase | TypeScript 97.8% |
| Latest release | v2026.416.0 (CalVer) |
| Runtime | Node.js 20+, pnpm 9.15+ |
| Database | PostgreSQL (embedded for dev, external for prod) |
| Frontend | React |
| Default API port | http://localhost:3100 |
Use cases: who should read the new docs
- First-time Paperclip users — v2 is the cleanest on-ramp yet. Start at the Getting Started guide, then run
npx paperclipai onboard --yes. - Adapter builders — if you're plugging in Claude Code, Codex, Cursor, OpenClaw, or a custom HTTP/webhook bot, the adapter docs got an explicit upgrade.
- Desktop users — pair the docs with Paperclip Desktop, the unofficial Electron wrapper (macOS today, Windows/Linux coming soon).
- Operators of autonomous companies — governance, budgets, and approval flows are exactly the pages where screenshots save you the most time.
- Contributors — a canonical reference before you open a PR against paperclipai/paperclip.
How to read v2
A pragmatic reading order for someone opening the docs cold:
- Start at Getting Started — you should have Paperclip running locally inside 5 minutes.
- Jump to Agents & Org Chart to understand the "company, not tool" mental model.
- Read Budgets & Governance before you ever let an agent spend real money.
- Skim the Adapter section for whichever runtime you plan to hire — Claude Code, Codex, Cursor, OpenClaw, or custom.
- Bookmark the Routines and Heartbeats pages for when you want recurring work.
Limitations & caveats
Two honest caveats worth pinning next to the launch tweet:
- Still community-maintained. v2 is not the official Paperclip project's documentation yet — Prins has said he hopes it becomes official soon, but there's no ETA.
- Not exhaustive. Prins himself flags "many more additions and improvements to be added soon," so don't be surprised if a specific edge case isn't covered yet. File issues on the docs repo if you find gaps.
What's next
Expect v3-ish incremental updates rolling out over the coming weeks as coverage fills in, and watch for a potential move to the official paperclipai org. In parallel, the core project keeps shipping (v2026.416.0 landed mid-April) and Paperclip Desktop's Windows/Linux builds are on the roadmap.
For now: if you've been eyeing Paperclip but bounced off the README, v2 of the docs is your best second-look. And if you find a gap, tell Prins — he's actively iterating.
Sources: Aron Prins on X, Paperclip Docs, paperclipai/paperclip, paperclip.ing.


