Paperclip

Manage agents as an organization with roles, reporting lines, budgets, and shared goals

The official README describes Paperclip as a Node.js server and React UI for orchestrating teams of AI agents. Agents can have roles, titles, reporting lines, permissions, and budgets, while company missions and goals provide context for downstream tasks. OpenClaw, Claude Code, Codex, Cursor, Bash, HTTP/webhook bots, and other heartbeat-capable runtimes can participate in the same org chart.

Sources: [1]

Put tasks, approval gates, auditability, and budget enforcement in one control plane

Paperclip's work system covers issues, dependencies, comments, documents, attachments, and work products, with atomic checkout and execution locks designed to avoid duplicate work. Approval workflows, pause/resume/terminate controls, revision history, token and cost tracking, warning thresholds, and hard stops give operators explicit control over autonomous execution.

Sources: [1]

Turn recurring agent work into an operating system of heartbeats, cron, webhooks, and API triggers

Agents can wake on heartbeats to inspect assigned work, while routines support cron, webhook, and API triggers. The README describes tracked issues for routine runs plus workspace resolution, secret injection, skill loading, adapter invocation, structured logs, cost events, session state, and audit trails around execution.

Sources: [1]

Use the Docker quickstart without installing Node.js or pnpm on the host

The official Docker guide recommends docker compose -f docker/docker-compose.quickstart.yml up --build. The UI opens at http://localhost:3100, while embedded PostgreSQL data, uploaded assets, the local secrets key, and agent workspace data persist under ./data/docker-paperclip. The image also includes Claude Code, Codex, OpenCode, and Gemini CLIs.

Sources: [2]

A strong fit when many coding or business agents run continuously and need one layer for ownership and control

Paperclip is most relevant when multiple Claude Code, Codex, or other agents are already running in parallel and terminal tabs or scattered configuration files no longer make ownership, progress, cost, and approvals clear. It is a comparison candidate when you want to manage agent roles, daily work, budgets, and human review from business goals rather than only launch isolated agent sessions.

More agent autonomy increases the importance of credential, workspace, budget, and persistent-data controls

The Docker quickstart can launch the application without provider API keys, but local adapters such as Claude Code, Codex, and Gemini require their respective credentials to perform agent runs. Treat secrets and workspace permissions as execution-environment security boundaries in addition to Paperclip's approvals, budgets, and audit controls.

./data/docker-paperclip contains embedded PostgreSQL data, uploaded assets, the local secrets key, and agent workspace data. Treat that directory as persistent operational state with an explicit backup, access-control, and disposal policy. The source is MIT licensed. For non-Docker development, package.json requires Node.js 24.11 or newer and pnpm 9.15.4.

Sources: [2][3][4]