paperclipai/paperclip
Paperclip has accumulated over 6,500 GitHub stars since its open-source launch in early 2026, establishing itself as the leading orchestration layer for coordinating teams of autonomous AI agents structured as companies. Built primarily in TypeScript (97.1%), the platform provides organizational infrastructure — org charts, hierarchical goal alignment, ticket-based task management, and per-agent budget enforcement — that treats AI agents not as isolated tools but as employees within a governed business structure. The architecture runs on a Node.js backend with PostgreSQL storage and a React-based dashboard UI. Developers can spin up a local instance with a single command (`npx paperclipai onboard --yes`), which boots an embedded PostgreSQL database and an API server on port 3100. For production deployments, Paperclip connects to an external PostgreSQL instance and can run on any Node.js hosting environment. The data model enforces company-scoped entity isolation, meaning a single deployment can manage multiple autonomous companies with complete data separation. Paperclip integrates with a broad range of agent runtimes including OpenClaw, Claude Code, Codex, Cursor, Bash scripts, and any HTTP-compatible system, avoiding vendor lock-in entirely. Agents wake on scheduled heartbeats, task assignments, or notifications, and their state persists across reboots and disconnections. The platform uses atomic task checkout combined with budget enforcement to prevent double-work and runaway token spending — agents auto-pause when they hit 100% of their monthly budget allocation. Governance features include immutable append-only audit logs with full tool-call tracing, board-level approval gates for strategic decisions and agent hiring, and revisioned configuration changes with rollback capability. Runtime skill injection via SKILLS.md files allows operators to modify agent capabilities without retraining. The project maintains an active roadmap including ClipMart, a planned marketplace for pre-built company templates.
Why It Matters
Running a single AI agent is straightforward. Running twenty agents simultaneously across different tasks, budgets, and objectives without losing track of what each one is doing, spending, and deciding — that is the coordination problem most teams hit once they move past proof-of-concept. Paperclip addresses this directly by providing the missing management layer between individual agent frameworks and actual autonomous operations. Before Paperclip, teams managing multiple agents relied on ad-hoc terminal sessions, manual cost tracking, and improvised task delegation with no audit trail. Nat Eliason's Felix agent demonstrated over $100,000 in autonomous revenue, and Aaron Sneed's 15-agent council saves 20+ hours weekly — but these results require the kind of structured orchestration that Paperclip provides. The platform's atomic budget enforcement solves the very real risk of a single misconfigured agent burning through thousands of dollars in API calls overnight. Its goal hierarchy ensures every task traces back to a company objective, so agents understand context rather than executing instructions blindly. For teams building production autonomous systems, Paperclip fills the gap between having capable agents and actually operating them at scale with accountability and cost control.