Paperclip
Paperclip Labs, Inc. · Open-Source Control Plane for Running Teams of AI Agents
Paperclip is a free, MIT-licensed, self-hosted app that runs agents you already use, such as Claude Code, Codex, Cursor, and OpenClaw, as a company: an org chart, goals, a monthly budget per agent, scheduled heartbeats, approval gates, and a ticketed audit trail. It is for builders running several agents at once who need cost control and oversight rather than another agent.
Best for
Technical founders, builders, and small teams already running several coding or general-purpose agents who want budgets, approvals, and an audit trail around them
Not ideal for
Anyone looking for a single agent to do work directly, non-technical users who want a hosted app with no setup, or developers who want a framework to build new agents in code
Who it's for
Builders, technical founders, and teams running multiple AI agents who need org structure, budgets, and governance around them
Capabilities
- Org chart for agents with roles, reporting lines, and delegation that routes work up and down the hierarchy
- Monthly budget per agent, with costs tracked per agent, task, project, and goal, and automatic pausing when an agent reaches its limit
- Scheduled heartbeats that wake agents to pick up assigned work, plus recurring routines
- Ticket system that records every instruction, response, tool call, and decision in an append-only audit log
- Board controls to pause, resume, override, reassign, or terminate agents, and approval gates for hiring agents and executing strategy
- Adapters for Claude Code, Codex, Gemini CLI, Cursor (local and cloud), OpenCode, Hermes, Pi, Grok, and Kimi Code, plus generic process and HTTP adapters for custom agents
- Connections store for AI runtime credentials (Claude and Codex subscriptions, API keys) and per-person GitHub identities
- REST API and CLI covering companies, agents, issues, approvals, costs, routines, secrets, and activity, with multiple isolated companies in one deployment
Limitations
- No hosted version is available yet: you run it yourself, the site offers a waitlist, and cloud deployments are listed as in progress on the roadmap
- Paperclip does not provide models, so real costs come from the subscriptions and API keys your agents use
- Requires Node.js 24.11+ and pnpm; the installation docs cover macOS and Linux, and Windows is not documented
- Chat connectors (Slack chat, Discord, Microsoft Teams, Telegram, iMessage), AgentMail, and the GitHub workflow connector are labeled experimental
- Not an agent framework: it coordinates agents you already have, so output quality depends on those agents
- Memory and knowledge, a desktop app, and using Asana, Linear, or Jira in place of Paperclip's own tickets are roadmap items, not yet available
Use cases
- Running Claude Code and Codex agents as an engineering team with a lead agent that breaks goals into tickets and delegates them
- Capping each agent's monthly spend so a looping agent pauses instead of burning through API credits
- Scheduling heartbeat-driven agents for recurring content, QA, or research work and reviewing their audit trails
- Requiring board approval before an agent hires another agent or executes a new strategy
- Giving agents their own AgentMail inboxes so incoming email becomes assigned work
- Hosting one Paperclip server on a VPS behind Tailscale to manage agents for several separate projects or companies
Paperclip pricing
Open source (self-hosted)
Free
- MIT license
- No Paperclip account required for local use
- Runs locally with embedded PostgreSQL or on your own server
Note: Paperclip publishes no paid plan. Costs come from the AI subscriptions or API keys your agents use and from any infrastructure you host Paperclip on.
Technical specs
Available models
Paperclip vs. competitors
Paperclip vs. Orca
Orca is a desktop environment for running coding agents side by side in git worktrees with you reviewing diffs; Paperclip organizes agents into a persistent org chart with goals, tickets, budgets, and scheduled heartbeats, and is not limited to coding work.
Paperclip vs. OpenHands
OpenHands ships its own coding agent plus a control center and automations for software work; Paperclip has no agent of its own and focuses on governance (budgets, approvals, audit trails) across many agents and business functions.
Paperclip vs. CrewAI
CrewAI is a framework for building role-based agent crews in code; Paperclip does not build agents but coordinates existing ones, such as Claude Code or Codex, as a governed company.
Paperclip vs. OpenClaw
OpenClaw is a personal agent; Paperclip can hire OpenClaw agents and manage them alongside others. The project describes it this way: if OpenClaw is an employee, Paperclip is the company.
Frequently asked questions
What is Paperclip?
Paperclip is an open-source app for managing teams of AI agents. It organizes agents you already use, such as Claude Code, Codex, or OpenClaw, into an org chart with goals, tickets, budgets, scheduled heartbeats, and approval gates, and records everything they do in an audit trail.
Is Paperclip free and open source?
Yes. Paperclip is MIT-licensed and self-hosted, and no Paperclip account is required for a local install. Your costs are the model subscriptions or API keys your agents use and any server you host it on.
Which agents work with Paperclip?
The docs list adapters for Claude Code, Codex, Gemini CLI, Cursor, OpenCode, Hermes, Pi, Grok, and Kimi Code, with an OpenClaw gateway and generic process and HTTP adapters for custom agents. Any agent that can receive a heartbeat can be connected.
Is there a hosted cloud version of Paperclip?
Not yet. Paperclip runs on your own machine or server, locally with embedded PostgreSQL or on a VPS with login. The website has a waitlist for a hosted version, and cloud deployments are listed as in progress on the project roadmap.
How do Paperclip's agent budgets work?
Each agent gets a monthly budget. Paperclip tracks costs per agent, task, project, and goal, and pauses an agent automatically when it reaches its limit, so a runaway agent cannot keep spending.
Integrations & fit
Alternatives to consider
Stably AI
Developers who already use one or more CLI coding agents and want to run and review many of them in parallel without branch conflicts
Free
OpenHands
Developers and platform teams who want an open-source, bring-your-own-model coding agent they can run and automate on their own machines or VMs
FreemiumCrewAI
Orchestrating autonomous agent teams for enterprise tasks
FreemiumAbout Paperclip
Paperclip does not build or replace agents; it manages the ones you already use. You set up a company with a mission and goals, hire agents into an org chart with roles and reporting lines, and assign work as tickets. Agents wake up on scheduled heartbeats, pick up their tasks, and delegate up or down the hierarchy, while recurring routines cover repeat work. Every instruction, response, tool call, and decision is recorded in an append-only log attached to the ticket, and you act as the board: you approve new hires and strategy, and you can pause, resume, override, reassign, or terminate any agent. Each agent has a monthly budget, costs are tracked per agent, task, project, and goal, and agents pause automatically when they hit their limit. Adapters cover Claude Code, Codex, Gemini CLI, Cursor, OpenCode, Hermes, Pi, Grok, and Kimi Code, and generic process and HTTP adapters connect custom agents, following the project's rule: if it can receive a heartbeat, it can be hired. The September 2026 release (v2026.916.0) moved AI credentials such as Claude and Codex subscriptions and API keys into a Connections store, gave each person their own GitHub identity, and added email inboxes for agents through AgentMail plus experimental chat connectors. Paperclip runs as a Node.js server and React UI with embedded PostgreSQL on your own machine, or on a server behind Tailscale or a public URL with login. The tradeoffs: there is no hosted version yet (the site has a waitlist, and cloud deployments are listed as in progress on the roadmap), it supplies no models, so your agents' subscriptions and API usage are the real cost, and it only adds value if you already run several agents worth coordinating.
Updates from Paperclip
A 503-commit release moved AI runtime credentials (Claude and Codex subscriptions, API keys, shared accounts) into Connections, gave agents dedicated AgentMail inboxes, replaced a shared GitHub token with per-person GitHub identities, and added experimental Slack, Discord, Telegram, Teams, and iMessage chat connectors.
Kimi Code became a first-class adapter, the company skill library now reaches agents at runtime, and onboarding was simplified.
Are you the founder? Claim this listing →