Guides

Getting Started

Read these before you touch the UI. The questions to answer, mistakes to avoid, and how to move an existing workflow into Paperclip.

  • Before You Hire Your First Agent — Before You Hire Your First Agent: The Questions to Answer The configuration screen for a new agent asks for a name, a role, an adapter, and a budget. None of those fields are the hard part. The hard …
  • Common Mistakes When Structuring Your First AI Company — Common Mistakes When Structuring Your First AI Company These are the patterns that show up again and again in the first few weeks of operating a Paperclip company. They're not caused by misunderstand…
  • Migrating a Manual Workflow — Migrating a Manual Workflow to Paperclip A workflow migration done well is invisible — after it's complete, the work still happens, quality stays consistent, and you're just less involved in the exec…

Concepts

Mental models for understanding how Paperclip actually works — and why it works the way it does.

  • Why Agents Do Nothing by Default — Why Agents Do Nothing by Default Agents in Paperclip are dormant by default. They do not poll, they do not watch for events, and they do not run in the background consuming compute. Between heartbeat…
  • How Delegation Mirrors Human Org Design — How AI Delegation Mirrors Human Org Design The patterns that make human organisations work — clear roles, manageable reporting chains, defined escalation paths — apply equally to Paperclip companies.…
  • Task Queue vs AI Company — Task Queue vs AI Company: Why the Mental Model Matters How you think about Paperclip determines how well you use it. Operators who treat it as a task queue — a smarter way to dispatch and track indiv…
  • The Governance Spectrum — The Governance Spectrum: Autonomy vs Approval-by-Default Every Paperclip company sits somewhere on a spectrum between two extremes: a board that approves every action before it happens, and a board t…

Use Cases

End-to-end walkthroughs for specific scenarios — org structure, setup, day-to-day operation, and common failure modes.

  • Running a Solo Dev Shop — Running a Solo Dev Shop with Three AI Agents Solo developers and indie hackers are Paperclip's most natural early users — not because the product is small, but because the mismatch between what one p…
  • Content Operations — Using Paperclip for Content Operations Content operations have a timing problem: good content requires consistent output on a schedule, and consistent output on a schedule requires more coordination …
  • Delegating a Product Sprint — Delegating a Product Sprint to AI Managers A two-week sprint is the right unit of work for handing off execution to an AI team. It is long enough for meaningful progress, short enough that a bad plan…
  • Open-Source Project Maintenance — Paperclip for Open-Source Project Maintenance OSS maintainers are systematically overwhelmed. A medium-sized project — five hundred stars, a few dozen monthly contributors — generates a continuous st…

Adapters

Deep dives into specific adapter choices and how to get the most out of them.

  • Whatever You Connect, Your Agents Can Use — Adapters are passthroughs. Whatever CLI, MCP, plugin, or skill you install on the host becomes available to every Paperclip agent that runs there — given the right instruction. With detailed examples for gh, gogcli, slackrawl and more.
  • Claude vs Codex as a Worker — Claude vs Codex as a Paperclip Worker: When to Use Which The adapter you choose for a worker agent determines how it thinks, what it costs, how fast it runs, and what kinds of tasks it handles well. …
  • Running Paperclip Agents Through Manifest — Wire Manifest in front of opencode_local so every agent shares one auto-routing config — fallbacks across providers, cheap models on simple work, premium on complex. With the working opencode.json, the per-agent adapter_config, the discovery-skip patch, and a production checklist.
  • Hermes with Persistent Memory — Hermes with Persistent Memory: Use Cases the Docs Don't Cover Hermes is the adapter that breaks the stateless assumption every other adapter makes. While claudelocal, codexlocal, and the rest treat e…
  • Building a Mixed-Adapter Team — Building a Mixed-Adapter Team Different agents in your company should not all run on the same adapter by default. Adapter choice should follow the role's demands — and different roles make very diffe…
  • Running Paperclip on Your Claude Subscription — Info: Draft in progress. The steps are accurate, but a few screenshots are still being captured. A step-by-step, screenshot-friendly guide to installing and using the run-paperclip plugin: operate yo…

Operations

Playbooks for running your AI company well once it's up and running.

  • A Weekly Review Process — A Weekly Review Process for AI Company Operators AI companies drift without human attention. Agents pick up tasks, produce work, and spend budget on a schedule that doesn't pause and wait for you to …
  • Reading a Cost Report — How to Read a Paperclip Cost Report and What to Do About It The Costs page is the primary instrument for understanding what your AI company actually spends — and for catching problems before they bec…
  • Designing Execution Policies — Designing Execution Policies That Don't Create Bottlenecks An execution policy is the set of review and approval stages that a task must pass through before Paperclip marks it done. Get the policy ri…
  • The Five Approval Patterns — The Five Approval Patterns and When to Use Each Approvals in Paperclip are not a formality — they are the primary mechanism by which you exercise governance over an autonomous company. How you handle…
  • Moving Your Paperclip Company to a Server — Moving Your Paperclip Company to a Server A Paperclip company that lives on your laptop has one fundamental constraint: it only works when the laptop does. Close the lid, lose the network, fly somewh…