Paperclip is an open-source app for managing teams of AI agents for work. People use it to give their agents an org chart, goals, tasks, and budgets instead of running a pile of disconnected automations. It suits operators who want agent work to stay visible, approved, and accountable inside one company workspace.[1][2]
Quick verdict
Teams that run several AI agents across business functions will get the most from Paperclip. Its org chart, goal and budget tracking, approval gates, and broad adapter and connector support cover the full assign-run-review lifecycle for agent work. Solo users who only need one personal assistant, and teams that want fully managed hosting today, should choose something else, since Paperclip is self-hosted software with a waitlisted cloud. If your staff must supervise agent teams with governance controls in infrastructure you operate, choose Paperclip; if you want a personal assistant over messaging with minimal setup, choose OpenClaw instead.[1][2][3][4]
Key features
Org chart with delegation paths Shape teams of agents, set reporting lines, and pick an adapter per agent instead of wiring automations from scratch.[1][2]
Goals, tasks, and work modes Track what each agent owes through projects, goals, issues, routines, and execution workspaces with a command palette and dashboard.[1]
Costs and budgets Watch spend per agent or team through dedicated cost tracking, so autonomous work stays inside a budget the operator sets.[1][2]
Approvals and activity log Gate consequential actions behind approvals, record decisions, and replay who did what in a timestamped activity log with a work timeline.[1][2]
Multi-provider agent adapters Run Claude, Codex, Gemini, Grok Build, Cursor, Kimi, or OpenCode agents, or a custom HTTP runtime, with per-message model and effort selection.[1][2][3]
Connector catalogue with access control Connect Gmail, Slack, GitHub, Notion, Linear, Google Workspace, and dozens more through one-click setup, MCP aggregators, and per-action permissions.[1][2]
Self-hosted deployment Run on a laptop, in Docker, or on a server with secrets management and Tailscale, with terminal setup and execution policies for power users.[1][2][3]
Control-plane API and CLI Operate companies, agents, issues, approvals, and costs programmatically through a 30-page API reference and a complete command-line reference.[1]
Who will love it
Operators running agents across business functions The org chart plus goals, tasks, budgets, and 75 pages of connectors fits dev, content, marketing, research, and outreach work in one supervised workspace.[1][2]
Teams that self-host their stack Local install, Docker, and server deployment with secrets and Tailscale keeps agent data and credentials inside infrastructure the team operates.[1][2]
Organizations needing oversight of autonomous work Approvals, decisions, blocked inbox handling, feedback voting, and the activity log give managers and boards a legible trail of what agents did and spent.[1][2]
Who should avoid it
Solo users wanting one simple assistant A person who only needs a chatbot over messaging will carry org-chart, governance, and connector machinery they never use.[1][2]
Teams wanting fully managed hosting now The vendor-hosted edition sits behind a waitlist, so the supported path today is installing and operating the local or self-hosted version.[1]
Non-technical buyers expecting turnkey setup Installation, model-provider connections, and connector configuration go through docs, CLI commands, and server administration, not a guided signup.[1][2]
Alternatives and compatibility
Against OpenClaw, choose the personal-assistant scope when one user driving agents over messaging is the whole job; OpenClaw is an open-source, MIT-licensed, self-hosted assistant with messaging integrations, while Paperclip adds the org layer of goals, budgets, approvals, and board-level controls. This comparison is editorial judgment from documented capabilities, not vendor endorsement.[1][2][3]
Against hosted agent platforms, choose the managed black box when nobody on the team will operate infrastructure or hold model keys.[1][2]
Teams already running container infrastructure get an extra incentive: Paperclip drops onto a laptop, Docker host, or Tailscale-connected server the team already understands.[1][2]
Questions worth asking
What is the best Paperclip alternative for non-technical users?
Is Paperclip really no-code?
How much does Paperclip cost per month?
The software is MIT-licensed and the local install carries no published price or per-seat fee; the waitlisted hosted edition has no published pricing. Model usage bills through the team's own provider connections, and hosting costs sit with the operator's infrastructure. Prices checked October 12, 2026.[1][2][3]
Sources and editorial review
Source research. Reviewed on October 12, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- Paperclip official homepage
official · Observed October 12, 2026
Identity anchor: product name, headline positioning as the app for managing AI agents for work, waitlist-or-local-install offering, org chart/goals/tasks/budget scope, v2026.1005.0 release notice.
- paperclipai/paperclip repository
official · Observed October 12, 2026
Repository identity paperclipai/paperclip, open-source app description, Dockerfile and docker directory, deployment and adapter doc references, MIT licensing references.
- Paperclip LICENSE (MIT)
official · Observed October 12, 2026
Exact license identifier MIT, copyright 2025 Paperclip AI.
- Paperclip docs index
official · Observed October 12, 2026
Guide coverage: quickstart, day-to-day operation, org and agents, 75 pages of connectors, projects and workflow, administration, API/CLI/adapters/plugins references, MIT footer.
- Paperclip deployment overview
official · Observed October 12, 2026
Laptop, Docker, and server deployment with secrets and Tailscale; CLI installation; Claude, Codex, Gemini, and custom HTTP adapter support.
- Paperclip v2026.1005.0 changelog
official · Observed October 12, 2026 · Published October 5, 2026
v2026.1005.0 dated October 5, 2026 with 179 commits: Grok Build on native runner, persistent agent directories, GitHub-synced skills, live browser, one-click connector setup, Slack round-trip, per-message model/effort picker, runner hardening.
- OpenClaw official site
official · Observed October 12, 2026
Competitor positioning: open-source AI assistant with product, install, ecosystem, integrations, docs, and community channels; supports the editorial competitor comparison scope.
Compared with these products
No explicit proprietary baseline is recorded.