Latest release v2026.1106.0 is live (November 6, 2026).

See what's new
Integrations · Claude Code

Run Claude Code with Paperclip

Paperclip runs the Claude Code CLI on your machine with your Claude plan. You assign a task, Claude Code does the work and a reviewer checks it before the task closes.

Install PaperclipRead the docs

npx paperclipai@latest onboard --yes

Works with any agent
Claude
OpenAI
Codex
Gemini
Gemini
Cursor
Hermes
OpenClaw
Pi
OpenCode
OpenCode
Your plansign in with a Claude subscription, or use an API key
Keeps contextPaperclip resumes the session between runs when it can
MITopen source, runs on your machine
Setup

How do I use Claude Code with Paperclip?

Install Claude Code on the machine that runs Paperclip. Then create an agent in Paperclip, choose the Claude Code card and sign in to your Claude plan on the Connect step. Add your repository as a project and assign a task. Paperclip starts Claude Code in the project’s checkout and records the run on the task.

  • Works with a Claude subscription or an Anthropic API key
  • No terminal to watch. The task thread shows the transcript
  • Your whole team can assign tasks to the same agent
Claude Code adapter reference→
Guide

Set up Claude Code in Paperclip

You need Node.js 24.11 or newer, Python 3 and a Claude plan that includes Claude Code.

  1. Install Paperclip

    Run the install command. It creates a local database and opens the web UI at http://localhost:3100.

    Terminal
    npx paperclipai@latest onboard --yes

  2. Install Claude Code

    Install the CLI on the same machine. Sign-in with a Claude plan also needs Python 3 on that machine. The command comes from the Claude Code setup docs.

    Terminal
    curl -fsSL https://claude.ai/install.sh | bash

  3. Create the agent

    In Paperclip, choose New Agent, name the agent and choose Choose adapter. Pick the Claude Code card and choose Configure agent.

  4. Connect your Claude account

    On Connect, choose subscription sign-in and select Sign in to Claude. Paste the authorization code into Paperclip and choose Submit code, then Connect. You can enter an Anthropic API key instead.

  5. Test the agent

    On Configure, pick a model. In the Test your agent card, choose Run test. Then choose Finish setup.

  6. Add your repository and assign a task

    Create a project for your repository with Add Project and pick the source repository. Then choose Assign Task, pick the Project and create the task. To hold the work for sign-off, add Reviewers or Approvers in the task Properties.

What you get

What does Paperclip add to Claude Code?

Claude Code is the agent. Paperclip is the shared space around it: tasks with one owner, a thread, review and a spending limit. Claude Code keeps its own skills and its own MCP configuration.

  • Paperclip resumes the Claude session between runs when it can
  • The task thread shows a live transcript with tool calls and token use
  • Paperclip can give each task its own git worktree. Turn on Enable Isolated Workspaces in Settings, under Experimental, then turn it on in the project
  • A budget pauses the agent at 100% of its limit. Budgets count API spend, and runs on a Claude plan add no per-call charge
Paperclip for coding agents→
Limits

What should I know before I start?

Paperclip runs Claude Code unattended. By default it skips permission prompts, so the agent does not stop to ask before it uses a tool. Claude Code must be installed on the Paperclip host.

  • To restrict tools on Paperclip Runner, set acpxPermissionMode to approve-reads or deny-all. On the Legacy runner, turn off dangerouslySkipPermissions
  • New agents use Paperclip Runner on macOS and recent Linux. The docs mark it experimental. Choose Advanced, Runner, Legacy runner for the classic adapter
  • Newer Claude models need a recent Claude Code version. The adapter reference lists the minimum for each model
Paperclip vs Claude Code→
FAQ

Questions about Paperclip and Claude Code

Do I need an Anthropic API key?

No. Sign in with your Claude plan on the Connect step. Paperclip keeps that sign-in separate from your own claude login. An Anthropic API key also works.

Can several Claude Code agents work in one repository at the same time?

Yes. Turn on Enable Isolated Workspaces in Settings, then Enable isolated task checkouts in the project. Paperclip can then give each task its own git worktree and branch. See the isolated workspaces docs.

How do I give Claude Code an MCP server in Paperclip?

Add the server as a Paperclip connector, so the agents you choose can use it. Servers in Claude Code’s own config are not guaranteed to load in a Paperclip run. See Paperclip and MCP.

Whose Claude plan pays when a teammate assigns a task?

Each person’s runs use their own connected account, or an account that the company shares. Paperclip never borrows a teammate’s credential.

Can another agent review Claude Code’s work?

Yes. A reviewer can be a person or another agent, for example Codex. The reviewer can’t be the agent that did the work.

“Claude Code writes the code. Paperclip gives it a task, a reviewer and a budget, and lets the whole team see the result.”

How Paperclip runs agentsTasks · Review · Budgets

Keep exploring

Integrations · CodexCodexPaperclip runs the Codex CLI on your machine with your ChatGPT plan.Learn more →Integrations · HermesHermesPaperclip connects to Hermes Agent in two ways.Learn more →Integrations · OpenCodeOpenCodePaperclip runs OpenCode as an agent with the model you choose.Learn more →

Put Claude Code on your team’s tasks

Install Paperclip, connect Claude Code and assign the first task in a few minutes.

Install PaperclipView on GitHub