Paperclip is an open-source orchestration platform designed to manage and run autonomous AI companies. Implemented as a Node.js server with a React UI, it provides a centralized control plane for coordinating diverse AI agents, enabling users to define business goals, assemble agent teams, and monitor their operations, costs, and progress. The core philosophy is to manage business objectives rather than low-level technical processes.

The system's methodology centers on several key architectural and operational principles:

1. Agent Agnosticism and Heartbeat Integration: Paperclip supports a "Bring Your Own Agent" model, compatible with any AI agent (e.g., OpenClaw, Claude, Codex, Cursor) capable of receiving a "heartbeat." This implies a standardized interface or API through which Paperclip can poll, activate, and assign tasks to agents. Agents are not confined to a specific framework but must adhere to Paperclip's orchestration protocol.

2. Hierarchical Goal Alignment: The platform establishes a structured hierarchy of goals, where every atomic task performed by an agent is traceable back to a high-level company mission. This "goal-aware execution" ensures agents consistently understand the "why" behind their actions, receiving full goal ancestry with each task assignment. This eliminates the need for manual context provision, as context flows automatically from company goals down to individual agent tasks.

3. Asynchronous and Persistent Agent Operation via Heartbeats: Agents operate based on scheduled "heartbeats" and event-based triggers (e.g., task assignment, @-mentions). This mechanism allows agents to wake up periodically, check for new work or status updates, and perform actions. Crucially, Paperclip ensures "persistent agent state," meaning agents can resume tasks from their last execution point across heartbeats, eliminating the need to restart processes or re-establish context from scratch.

4. Atomic Execution and Budget Enforcement: To prevent redundant work and control costs, Paperclip implements "atomic execution" for task checkout and budget enforcement. When an agent picks up a task, it's atomically assigned, preventing other agents from duplicating the effort. Concurrently, "cost control" mechanisms are in place, tracking token usage or API calls against pre-defined monthly budgets per agent. Upon hitting budget limits, agents are automatically throttled or stopped, preventing "runaway loops" and ensuring financial accountability.

5. Organizational Structure and Governance: Paperclip models companies with explicit "Org Charts," defining hierarchies, roles, and reporting lines for AI agents. This structured delegation allows agents to have assigned responsibilities and "bosses." Governance features provide human oversight, enabling management to approve agent hires, override strategies, pause, or terminate any agent. A "Ticket System" records every conversation, decision, and tool-call tracing in an immutable audit log, facilitating transparency and post-hoc analysis. Configuration changes are versioned, and rollbacks are supported for safe operation.

6. Runtime Skill Injection and Portable Templates: The system supports "runtime skill injection," allowing agents to dynamically learn Paperclip workflows and project context without requiring retraining. Furthermore, it introduces "portable company templates" (via "Clipmart" - a planned feature), enabling the export and import of entire organizational structures, agent configurations, and skills, with provisions for secret scrubbing and collision handling.

7. Multi-Tenancy and Isolation: Paperclip is designed for "multi-company" operation, allowing a single deployment to manage an unlimited number of companies with complete data isolation and separate audit trails, acting as a control plane for a portfolio of autonomous businesses.

In essence, Paperclip transforms the ad-hoc management of individual AI agents into a structured, governed, and cost-controlled business operation by providing an enterprise-grade orchestration layer that models real-world company dynamics.