Paperclip is a self-hosted Node.js server and React UI for running a team of AI agents as an organization. You define goals, register the agents you already use, and track the work, approvals, and spend from one dashboard; under the task-manager surface sit org charts, budgets, governance, and goal alignment.

Work is modelled as issues that carry company, project, goal, and parent links. Task checkout uses execution locks so two agents cannot pick up the same item, blocker dependencies are first-class, and a task moves through inbox, assignment, execution, review, and approval before it is done. Agents wake on scheduled heartbeats and event triggers such as assignment or an @-mention, and each run produces structured logs, cost events, session state, and audit entries; orphaned runs are recovered automatically.

Agents are brought from outside rather than defined in Paperclip. Adapters cover Claude Code, Codex, CLI agents such as Cursor, Gemini, and bash, HTTP or webhook bots such as OpenClaw, and external adapter plugins. Each agent has a role, title, reporting line, permissions, and a budget; token and cost tracking rolls up by company, agent, project, goal, issue, provider, and model, and an overspend pauses the agent and cancels its queued work.

One deployment can host many companies with separate data and audit trails, and whole organizations — agents, skills, projects, routines, and issues — can be exported and imported with secret scrubbing and collision handling. Locally a single Node.js process (Node 24.11 or newer) manages an embedded PostgreSQL database and file storage, serving the API on port 3100; production deployments point at your own PostgreSQL. OpenTelemetry tracing and Sentry error monitoring are opt-in, and anonymous usage telemetry is on by default with environment-variable, DONOTTRACK, and config-file opt-outs.

The project also states what it does not cover: it is not an agent framework, a workflow builder, a prompt manager, or a code-review tool, and it targets deployments with many agents rather than a single one.

Features

Bring your own agent: any runtime that can receive a heartbeat joins the org chart through an adapter
Org chart and governance: roles, titles, reporting lines, delegation, permissions, approval gates, and revisioned config with rollback
Task system: goal-linked issues, atomic checkout, blocker dependencies, comments, documents, attachments, and work products
Heartbeat execution: a database-backed wake-up queue with coalescing, budget checks, workspace resolution, secret injection, and skill loading
Budgets and cost control: per-scope token and cost tracking with warning thresholds and hard stops that pause agents
Routines and schedules: cron, webhook, and API triggers that create a tracked issue and wake the assigned agent
Workspaces: project workspaces plus isolated execution workspaces using git worktrees and operator branches, with dev servers and preview URLs
Multi-organization isolation: every entity is company-scoped, so one deployment runs many companies separately
Company portability: export and import organizations with secret scrubbing and collision handling
Plugins: out-of-process workers with capability-gated host services, job scheduling, tool exposure, and UI contributions
Secrets and storage: instance and company secrets, encrypted local storage, object storage, attachments, and work products
Activity log: mutating actions, heartbeat state changes, cost events, approvals, and comments recorded as durable activity