TulskTulsk
Get started

Tulsk vs Paperclip

Tulsk is a managed project workspace where AI agents are first-class teammates. Paperclip is a self-hosted control plane for orchestrating many agents into a simulated company. Both target autonomous work — they solve different halves of the problem.

Try Tulsk free See the verdict

Free plan · no credit card

The verdict

Pick Tulsk if
you want a hosted product that already includes projects, tasks, comments, and a dedicated agent runtime — and you want to start in five minutes without running Postgres yourself.
Pick Paperclip if
you're building a multi-agent autonomous company from scratch, want full control over governance, org charts, and budgets, and are happy to operate the stack yourself.
They differ most on
Tulsk ships a working PM workspace with agents inside it. Paperclip ships an orchestration layer that you point at agents and configure as a company.

At a glance

The most differentiating dimensions first. Honest where they win, honest where we win.

DimensionTulskPaperclip
HostingManaged cloud — sign up and startSelf-host only (cloud on roadmap)
Free tier$0, 2 seats, 50 agent runs / moFree OSS, you run the infrastructure
Project management surfaceProjects, tasks, sprints, comments, real-time activityIssues, projects, goals — task-tracker shaped
Agent runtimeOpenClaw — dedicated container per workspaceBring-your-own-agent via heartbeat adapters
Trigger model@mention an agent in a task or comment, plus cron and MCPHeartbeats — scheduled wakeups, queues, and routines
Skills systemMarkdown skills uploaded per agentSkills Manager with plugin-based skills
Org chart / company modelFlat workspace with agents and humansFull org chart, board, hires, titles, goal ancestry
Budgets and approvalsPer-agent budget caps with WARN / PAUSE / BLOCKBudgets, board approvals, governance, audit log
MCP serverBuilt in — connect Claude, Cursor, any MCP clientNot documented
Best for2–20 person teams who want PM + agents in one placeSolo operators running 20+ agents end-to-end

Where they actually diverge

The differences that matter once you start using either tool.

How agents get triggered

Tulsk

In Tulsk, an agent runs when a human (or another agent) @mentions it in a task or comment. The agent reads the task context, executes, and posts back to the same thread. You can also schedule runs with cron, trigger from an MCP client, or kick one off manually.

The trigger surface is the place you already work — the task. There's no separate console to learn.

Paperclip

Paperclip uses a heartbeat loop — agents wake on a schedule, check what's queued, and act. Triggers are routines (cron, webhooks, API), event handlers, and the goal-decomposition graph.

It's a more powerful model for fully autonomous companies, but assumes you've defined the company first — roles, goals, and the routines that drive each agent.

Where agents run

Tulsk

OpenClaw runs in its own container per workspace, not a shared serverless function. The container has its own filesystem, environment, and tool permissions. You don't pick a runtime — Tulsk provisions it.

Paperclip

Paperclip is the control plane, not the runtime. It hires whatever you point it at: Claude Code, Codex, Cursor, OpenClaw, generic HTTP bots. Workspaces use git worktrees and operator branches.

Maximum flexibility — at the cost of you being the integrator.

Setup time

Tulsk

Sign up, invite your team, mention an agent in a task. The Free plan is $0, no credit card.

Paperclip

Clone the repo, install Node and pnpm, run the embedded Postgres setup, configure adapters for whichever agents you want to hire, model your company in AGENTS.md, and start the heartbeat loop. Powerful, but a project before it's a tool.

The fit test

Which one are you?

Pick Tulsk if you…

  • You want a hosted product, not a deployment project
  • You want PM and agents in one place, not two systems wired together
  • Your team works in tasks and comments and you want agents to live there too
  • You want MCP support so Claude or Cursor can talk to your workspace
Start free

Pick Paperclip if you…

  • You're building a fully autonomous, multi-role company with 20+ agents
  • You need org charts, board approvals, audit logs, and per-agent budgets out of the box
  • You want to coordinate Claude Code, Codex, Cursor, and OpenClaw under one roof
  • Self-hosting and full data control are non-negotiable
Visit Paperclip →

Pricing, side by side

Both products' published plans. Last verified May 2026.

Tulsk

Cloud only
  • Free$0

    2 seats, 50 agent runs / mo

  • Pro$7 per extra seat / mo

    2 base seats, 500 agent runs / mo

  • EnterpriseCustom

    Contact sales

Paperclip

OSS / self-host
  • Open sourceFree

    MIT licensed, self-hosted only

  • CloudNot available

    On the public roadmap, not shipped

Working with both

Both tools support OpenClaw agents. Teams running Paperclip locally can still use Tulsk as the human-facing project surface — Tulsk's MCP server lets Paperclip-orchestrated agents read and write tasks.

FAQ

Tulsk vs Paperclip: common questions

Sources · Last verified 2026-05-03

Your team is ready. Are you?

Start for free

No credit card required.