Munder Difflin releases open-source harness that wraps Claude Code to build 'your clone'

It all starts with one download file

Just install one download file on your laptop, and that's it. It wraps around whatever coding agent CLI you're already using and layers on top of it a "clone" modeled after you. According to the project, your code, API keys, and existing subscriptions stay put — nothing gets sent outside. A post introducing the project hit Hacker News on the 22nd and drew more than 200 upvotes.

The clone absorbs your workflow, your tools, and the knowledge you've built up over time. That means every new clone you spin up already understands how you work from the moment it starts. The GitHub repository is public under the account chaitanyagiri, and you can grab the download from the releases page.

Why a harness, and why now

A "harness" isn't the AI model itself — it's the framework that plugs a model into a real working environment. A recent example that captures this trend well: NVIDIA wrapped its own harness around Claude Opus 5 and pushed benchmark scores for the same model from the 30% range up to 100%. Increasingly, how you wrap a model matters as much as the model itself when it comes to real-world performance.

Around the same time, AWS released a bridge that lets cloud-based agents reach tools on a user's local machine, and Cursor rolled out 31 official plugins connecting business tools like Gmail and Salesforce to its agents. Munder Difflin sits in that same wave, but takes a different approach: rather than building one shared bot for the whole team, it builds a separate clone for each team member.

How the clones actually work

Each clone is a node running on that person's own machine. When someone's clone hits a wall mid-task, it messages a teammate's clone to get the information it needs. The project gives an example: if a colleague's clone sends a question like "How does the billing service work?" at 3 a.m., your clone answers on your behalf without waking you up.

Clones review teammates' pull requests according to your own standards and habits, fix bugs, watch CI pipelines, and keep documentation up to date — carrying out their assigned roles around the clock. The idea is that only the handful of decisions that genuinely require human judgment get escalated to you, while everything else gets handled clone-to-clone.

How to try it

① Where to start — Download the node file from the releases page of the github.com/chaitanyagiri/munder-difflin repository.

② Step by step — First, pick a CLI agent you're already using — Claude Code, Codex, Cursor, or similar — and connect it, carrying over your existing API key or subscription account as-is. Then run the node to create a clone that learns how you work, and if you have teammates, you can set it up to exchange messages with their clones.

③ Who can use it — The personal node is free and MIT-licensed, so you can inspect the code directly on GitHub. That said, running a clone still requires an existing subscription or API key with a provider like Claude, OpenAI, or Copilot. To let clones message each other and share a common knowledge base across a team, you'll need a Teams Lite or PRO license, with seat tiers scaling from 10 seats up to 100-plus.

④ What you can actually do with it — For example, a designer's clone can cross-check screens against the design system to catch issues and pull out assets, while a product manager's clone can organize issues and keep boards and docs in sync. A salesperson's clone can draft outreach and keep the CRM current. Essentially, the pitch is that almost anything you can do from a command line on your computer can be handed off to a clone.

Pricing structure

Tier	Where it runs	Clone-to-clone messaging	Knowledge base
Personal (free, MIT)	User's laptop	Not supported	Personal only
Teams Lite	User's laptop	End-to-end encrypted	Shared team knowledge base
Teams PRO	Dedicated sandbox VM per clone	End-to-end encrypted	Shared team knowledge base
Cloud + Network	Dedicated sandbox VM, runs 24/7	End-to-end encrypted	Hosted organization-wide knowledge base

If you want your clone to keep running even after you close your laptop, you can move it to a dedicated sandbox VM under the Cloud + Network license — and the project says you can always move it back to local later.

Privacy and security design

The project's stance is that for clones to be trusted, users need control over where they run and who can read their messages. Messages exchanged between clones are encrypted on the sending node and only decrypted on the receiving teammate's node — the developers say even they can't see the contents in transit. Users can decide for themselves what knowledge is shared across the whole team versus what stays private context for just them, and once a shared knowledge base is set up, it's version-controlled and carries over to any new clones created afterward.

Editor's take

What makes this project interesting is that it flips the default assumption that's dominated so far: one shared bot for the whole team. Most agents companies have deployed until now have been a single chatbot split across many users — even the finance-team agent AWS recently unveiled was a single instance the whole organization logs into. Munder Difflin goes the other way, creating one clone per team member and letting those clones talk to each other. It's essentially porting the human org chart directly into software.

Whenever you actually put a harness layer like this together, you hit the same sticking point every time — not model performance, but deciding whose context gets shared, and how much. Without a clear line between a person's private storage and the team's shared knowledge base, there's a real risk that a clone leaks sensitive code or customer data to another clone without the team member ever knowing. How tightly a tool like this locks that down with encryption and permission separation is really the true measure of its quality.

For small dev teams or startups in Korea, the safer path is to start with the free personal version and let one or two people test it locally first. Hand off repetitive, clearly-rule-bound work like PR reviews or doc syncing to the clones first, and keep humans in the loop for anything involving sensitive data — customer information, contracts — until you're ready to move up to a Teams license. In the coming weeks, it wouldn't be surprising to see companies that already control CLI ecosystems, like Cursor or GitHub Copilot, try to build similar "clone messaging" features into their own products.