Paperclip: Orchestrate AI Agents as Employees for Zero-Human Ops
Run `npx paperclip-ai onboard` to create an org chart of AI agents using any LLM; assign tasks via CEO agent, enforce QA/approvals, and automate routines to handle marketing, coding, or sales without coding skills.
Build AI Organizations via Simple Org Charts and CEO Delegation
Start Paperclip with npx paperclip-ai onboard to spin up a CEO agent powered by Claude, Codex, Gemini, or cheaper options like Qwen 2.5 via OpenRouter. Act as the human CEO: assign high-level goals like "Hire a video writer and create a Remotion animation celebrating 40k GitHub stars." The CEO breaks tasks down, hires specialists (e.g., CTO with coders), installs skills (e.g., Remotion best practices from skills.sh), and executes. Agents access shared context like brand guides and dashboards automatically, producing on-brand outputs in minutes—what might take a week manually. Build incrementally: approve hires/plans agent-by-agent to ensure quality; avoid bloated templates with 130+ agents. Set per-agent/project budgets and track monthly spend (starts at $0 using subscriptions). Customize instructions iteratively, e.g., "If blocked, provide fix tutorial; write partial tests only," to refine behavior. Use meta-agents like skill consultants to diagnose skill underuse across the org.
Enforce Reliability with Workflows, QA, and Routines
Mandate workflows for completion: tasks auto-route to QA agents (using agent-browser skill for site testing/forms) post-assignee, enabling coder-QA iteration before manager approval. This vendor-neutral harness works across LLMs, preventing issues like agents skipping browser tests in multi-tab chaos. Routines automate repeats: schedule/group tasks like "Post Discord on master branch merges," "Generate changelog," or "Review PRs with Greptile skill." Parameterize with templates (e.g., branch name) for reuse over prompt folders. Agents persist memory/conversations, learning preferences like 2-second video cuts vs. 6-second defaults, evolving generic skills into org-specific ones (e.g., Paperclip branding/pacing). Non-coders handle marketing/sales/finance; coders pair with Cursor/GitHub Copilot externally.
Scale to Production Businesses with Upcoming Automation
Paperclip (released March 4, hit 50k GitHub stars by April 8 in 34 days) supports parallel execution (default 1, configurable), experimental workspaces for PRs/worktrees, and future CEO chat for natural delegation. Roadmap adds: maximizer mode (burn tokens relentlessly), multi-human teams/cloud deploys, sandboxing (E2B/devbox), desktop app, artifacts/deployments, action buttons on reports (e.g., "Create issue/plan"), auto-org learning. Start small for any business—e.g., proxy SaaS tools directory: CEO hires CTO, prototypes integrations/CI in minutes. Humans control chaos: debug, guide, inject taste for brand-quality output, commanding thousands of agents to amplify leverage.