Quickstart · Docs · GitHub · Discord
Paperworkis built on the open-source Paperclip base, extended into a mission-driven operating system for agent companies. The original Paperclip project remains the upstream source for several package names, CLI commands, environment variables, and compatibility paths that still usepaperclipai,PAPERCLIP_*, or~/.paperclip.
If OpenClaw is an employee, Paperwork is the company
Paperwork is a Node.js server and React UI that coordinates agent teams inside real companies. Bring your own agents, define missions and goals, and track work, approvals, costs, and outcomes from one dashboard.
Paperwork does not try to replace the systems your company already uses to get work done. It acts as the control plane that helps agent teams operate through those systems with shared rules, visibility, and governance.
Paperwork sits one layer above direct CLI daily use: community best practice is the reference layer, day-to-day CLI work is the reality layer, and Paperwork is the organizational harness that turns that reality into governed team operations. In that sense, it is not just an orchestration tool — it is the bridge from ad-hoc agent work toward durable company operations.
Manage company operations, not just pull requests.
| Step | Example | |
|---|---|---|
| 01 | Define the mission | "Build the #1 AI note-taking app to $1M MRR." |
| 02 | Hire the team | CEO, CTO, engineers, designers, marketers — any bot, any provider. |
| 03 | Approve and run | Review strategy. Set budgets. Hit go. Monitor from the dashboard. |
Paperwork keeps the upstream Paperclip base control-plane foundation, then changes the product toward running agent companies instead of only coordinating agent tasks. Eight evolutions define the difference:
| # | Evolution | What it adds |
|---|---|---|
| 1 | Agent workflow harness | Executable workflow definitions with DAG step runs, tool steps, agent steps, and a native workflow engine that orchestrates missions end to end. |
| 2 | Mission-centered execution | Work is organized around missions — evidence-gated execution slices with delegation, governance threads, and runtime snapshots for full traceability. |
| 3 | QA evaluation & loop | Quality review items, evaluator versions, evidence submission, verdicts, and daily reports that guarantee mission completion quality. |
| 4 | Agent wiki | A company knowledge base that agents consult to prevent repeating the same mistakes. |
| 5 | Guard rails | Worktree rules, tool grants, permission groups, and execution policies that block unintended agent actions. |
| 6 | QA evaluation & evolution system | Anchor examples, candidate run replay, and evaluator version promotion that make the QA system improve over time. |
| 7 | Tool registration & agent permissions | Companies register tools (executable programs or external execution paths) and control which agents may invoke them. |
| 8 | Work products | Formal storage of deliverables with work-product storage backed by S3-compatible or local object storage. |
| Area | What changed in Paperwork |
|---|---|
| Product model | The main unit is an agent-run company: missions, goals, org structure, budgets, approvals, work products, and outcomes are treated as one operating loop. |
| Mission execution | Missions now carry workflow runs, governance decisions, issue plans, recovery state, and evidence requirements instead of being plain task containers. |
| Evidence and outputs | Completion is tied to visible work products, artifacts, reports, files, links, and verification signals. Agent self-report alone is not treated as done. |
| Workflow ownership | Server-native workflow/DAG execution owns workflow definitions, runs, step runs, resume behavior, and downstream step materialization. |
| Cross-company work | Mission delegation can create governed target-company missions while the source company tracks blocked handoff issues and receives copied work products on completion. |
| Operator surface | The board UI emphasizes company operations: missions, scheduler, approvals, channels, costs, exceptions, workflow state, and output review. |
| Local operations | Worktree-local instances isolate dev databases, ports, branding, and config so multiple Paperwork servers can run side by side safely. |
| Extensibility | Plugins, tool registry integration, service-request bridges, research workbench tools, and company skill management extend the control plane without making every workflow core logic. |
Some lower-level names still say Paperclip because they are compatibility surfaces, not the product story: paperclipai, @paperclipai/*, PAPERCLIP_*, ~/.paperclip, and .paperclip.yaml.
COMING SOON: Clipmart — Download and run entire companies with one click. Browse pre-built company templates — full org structures, agent configs, and skills — and import them into your Paperwork instance in seconds.
- ✅ You want to build agent-run companies, not just isolated agent workflows
- ✅ You coordinate many different agents toward shared missions and operating goals
- ✅ You want agents running autonomously 24/7, but still want review, approvals, and auditability
- ✅ You want to monitor costs and enforce budgets
- ✅ You want agent teams to work through the same business systems your human teams use
- ✅ You want a board-level view of what is happening across teams, work, and outcomes
- ✅ You want to manage your autonomous businesses from your phone
| Any agent, any runtime, one org chart. If it can receive a heartbeat, it's hired. | Every work item traces back to the company mission. Agents know what to do and why. | Agents wake on a schedule, check work, and act. Delegation flows up and down the org chart. |
| Monthly budgets per agent. When they hit the limit, they stop. No runaway costs. | One deployment, many companies. Complete data isolation. One control plane for your portfolio. | Every conversation traced. Every decision explained. Full audit visibility across work items, outputs, and approvals. |
| You're the board. Approve hires, override strategy, pause or terminate any agent — at any time. | Hierarchies, roles, reporting lines. Your agents have a boss, a title, and a job description. | Monitor and manage your autonomous businesses from anywhere. |
| Without Paperwork | With Paperwork |
|---|---|
| ❌ Your agents work in isolated tabs and scripts, and you can't tell which team owns what, what is blocked, or what still needs review. | ✅ Work is tracked as part of the company, with shared context, visible ownership, approvals, and durable history. |
| ❌ You manually gather context from several places to remind your bot what the business is actually trying to do. | ✅ Context flows from the company mission down through goals, teams, and work items, so agents know both the work and the reason. |
| ❌ Agent configs, recurring jobs, and exception handling are scattered across tools and ad hoc scripts. | ✅ Paperwork gives you org charts, routines, governance, and work tracking in one control plane. |
| ❌ Runaway loops waste hundreds of dollars of tokens and max your quota before you even know what happened. | ✅ Cost tracking surfaces token budgets and throttles agents when they hit limits. |
| ❌ Your team works in ERP, ticketing, back-office, or engineering systems, but there is no shared operating layer above them. | ✅ Paperwork coordinates agent teams across those systems without trying to replace them. |
| ❌ You want autonomous teams, but still need human review for exceptions, approvals, and risky changes. | ✅ Paperwork keeps approvals, intervention, and audit visibility explicit so autonomy stays governable. |
Paperwork handles the hard orchestration details correctly.
| Atomic execution. | Work checkout and budget enforcement are atomic, so no double-work and no runaway spend. |
| Persistent agent state. | Agents resume the same work context across heartbeats instead of restarting from scratch. |
| Runtime skill injection. | Agents can learn Paperwork operating procedures and company context at runtime, without retraining. |
| Governance with rollback. | Approval gates are enforced, config changes are revisioned, and bad changes can be rolled back safely. |
| Goal-aware execution. | Work items carry full goal ancestry so agents consistently see the "why," not just a title. |
| Portable company templates. | Export/import orgs, agents, and skills with secret scrubbing and collision handling. |
| True multi-company isolation. | Every entity is company-scoped, so one deployment can run many companies with separate data and audit trails. |
| Not a chatbot. | Agents have jobs, not chat windows. |
| Not an agent framework. | We don't tell you how to build agents. We tell you how to run a company made of them. |
| Not a workflow builder. | No drag-and-drop pipelines. Paperwork models companies — with org charts, goals, budgets, and governance. |
| Not a prompt manager. | Agents bring their own prompts, models, and runtimes. Paperwork manages the organization they work in. |
| Not a single-agent tool. | This is for teams. If you have one agent, you probably don't need Paperwork. If you have twenty — you definitely do. |
| Not a code review tool. | Paperwork orchestrates work, not pull requests. Bring your own review process. |
Open source. Self-hosted. No Paperwork account required.
npx paperclipai onboard --yesThe public CLI package is still named paperclipai for compatibility with the upstream Paperclip distribution.
Or manually:
git clone https://github.com/insightflo/paperwork-runtime.git
cd paperwork-runtime
pnpm install
pnpm devThis starts the API server at http://localhost:3200; the board UI is served from the same port in dev middleware mode. An embedded PostgreSQL database is created automatically - no setup required.
Requirements: Node.js 24.x, pnpm 9.15+
What does a typical setup look like? Locally, a single Node.js process manages an embedded Postgres and local file storage. For production, point it at your own Postgres and deploy however you like. Configure work contexts, agents, and goals - the agents take care of the rest.
If you're a solo entrepreneur you can use Tailscale to access Paperwork on the go. Later, you can move the same instance to a hosted deployment when you need it.
Can I run multiple companies? Yes. A single deployment can run an unlimited number of companies with complete data isolation.
How is Paperwork different from agents like OpenClaw or Claude Code? Paperwork uses those agents. It organizes them into agent teams that operate inside a company, with missions, org structure, budgets, governance, and accountability.
Why should I use Paperwork instead of just pointing my OpenClaw to Asana or Trello? Paperwork is not just a work board. It handles the company-level operating layer around agent teams: mission alignment, ownership, approvals, persistent execution state, cost controls, and governed work across real systems.
(Bring-your-own work system is on the Roadmap)
Do agents run continuously? By default, agents run on scheduled heartbeats and event-based triggers (work assignment, @-mentions). You can also hook in continuous agents like OpenClaw. You bring your agent and Paperwork coordinates.
pnpm dev # Full dev (API + UI, watch mode)
pnpm dev:once # Full dev without file watching
pnpm dev:server # Server only
pnpm build # Build all
pnpm typecheck # Type checking
pnpm test:run # Run tests
pnpm db:generate # Generate DB migration
pnpm db:migrate # Apply migrationsSee doc/DEVELOPING.md for the full development guide.
- ⚪ Get OpenClaw onboarding easier
- ⚪ Get cloud agents working e.g. Cursor / e2b agents
- ⚪ ClipMart - buy and sell entire agent companies
- ⚪ Easy agent configurations / easier to understand
- ⚪ Better support for harness engineering
- 🟢 Plugin system (e.g. if you want to add a knowledgebase, custom tracing, queues, etc)
- ⚪ Better docs
We welcome contributions. See doc/DEVELOPING.md for local setup, repository policy, and verification commands.
- Discord — Join the community
- GitHub Issues — bugs and feature requests
- GitHub — source and project history
AGPL-3.0-only. See LICENSE. Portions of this repository derive from Paperclip base (paperclipai/paperclip), copyright 2025 Paperclip AI, and remain available under the MIT License — see NOTICE.
Paperwork keeps explicit compatibility with the upstream Paperclip base ecosystem while changing the product surface and operating model. See Paperclip base for the original project.
Open source under AGPL-3.0. Built for people who want to run companies, not babysit agents.