llms.txt
# Munder Difflin
> Munder Difflin gives you clones for you and your team, working 24/7. It is a free,
> MIT-licensed, local-first multi-agent harness for CLI coding agents that turns Claude Code,
> OpenAI Codex, Antigravity (Gemini), GitHub Copilot CLI, xAI Grok, Kimi Code, Qwen, OpenCode,
> Crush, Gemini CLI, Cursor Agent, and pi.dev into a coordinated "office" of agents on
> your own computer — each agent a real CLI process in a real terminal, with shared
> long-term memory, inter-agent messaging, your clone ("Michael") orchestrating, scheduled missions, Slack
> and webhook triggers, realtime voice orchestration, a built-in Monaco IDE with git
> history and branch compare, markdown previews, and auto-update. The interface runs in
> English, Simplified Chinese and Arabic, with right to left layout when Arabic is chosen,
> and its fonts are bundled rather than fetched from a font host. Current version: 0.4.6
> (August 2026). Desktop app for macOS, Windows, and Linux. No vendor cloud; code never
> leaves your machine. It covers the same jobs as YC's qm (security postures, background
> crons/watches, Slack, shared skills) as a simpler local-first desktop app, and adds what
> a headless harness doesn't have: watchable terminals, voice, an IDE, auto-update.
## Product
- [Homepage](https://munderdiffl.in/): features, download links, FAQ
- [GitHub repository](https://github.com/chaitanyagiri/munder-difflin): source, issues, releases
- [Latest release](https://github.com/chaitanyagiri/munder-difflin/releases/latest): macOS dmg, Windows setup, Linux AppImage
- [Changelog](https://github.com/chaitanyagiri/munder-difflin/blob/main/CHANGELOG.md)
## Key guides
- [What's up with Munder Difflin](https://munderdiffl.in/blog/whats-up-with-munder-difflin/): everything shipped from v0.3.3 to v0.3.7
- [Why our auto-update never ran](https://munderdiffl.in/blog/why-our-auto-update-never-ran/): the v0.3.7 fix — a CommonJS export that vanished into an ESM namespace
- [What's new in v0.3.5](https://munderdiffl.in/blog/launching-munder-difflin-v0-3-5/): the v0.3.4+v0.3.5 release wave
- [How to install and use Munder Difflin](https://munderdiffl.in/blog/how-to-install-and-use-munder-difflin/)
- [Your first hour with Munder Difflin](https://munderdiffl.in/blog/your-first-hour-with-munder-difflin/)
- [Frequently asked questions](https://munderdiffl.in/blog/munder-difflin-faq/)
- [How the GOD orchestrator works](https://munderdiffl.in/blog/how-the-god-orchestrator-works/)
- [How to talk to your orchestrator (voice)](https://munderdiffl.in/blog/how-to-talk-to-your-orchestrator/)
- [How to use the built-in Monaco IDE](https://munderdiffl.in/blog/how-to-use-the-built-in-monaco-ide/)
## Comparisons
- [qm vs Munder Difflin](https://munderdiffl.in/blog/qm-vs-munder-difflin/): YC's multiplayer harness vs a local-first agent office
- [Claude Squad vs Munder Difflin](https://munderdiffl.in/blog/claude-squad-vs-munder-difflin/)
- [Best Claude Code multi-agent tools](https://munderdiffl.in/blog/best-claude-code-multi-agent-tools/)
<!-- BLOG:START generated by blog/scripts/postbuild.mjs -->
## Blog
212 posts, grouped by topic, newest first in each topic. Topic hubs: https://munderdiffl.in/blog/topics/
### concepts
- [Google Let Gemini 4 Argon Loose on Its Own Coding Backlog. Here Are the Receipts.](https://munderdiffl.in/blog/gemini-4-argon-inside-google/): Google's own numbers for Gemini 4 Argon on real code: 134 bugs triaged in under 3 hours, 30% fewer rollbacks, a framework tuned overnight. Checked 13 Oct 2026.
- [Gemini 4 Argon: Google's Flagship Model Shares a Name With a Gas That Barely Reacts](https://munderdiffl.in/blog/gemini-4-argon/): Gemini 4 Argon went on general release on 13 Oct 2026: $2 input and $10 output per 1M tokens through 30 Nov, a 1M token window, and where you can use it.
- [Muse Wonder and Muse Code: Meta's new models, prices and scores](https://munderdiffl.in/blog/muse-wonder/): Meta's Muse Wonder and Muse Wonder Full: API price, Muse Code plans, context window and Meta's own coding scores, checked 13 Oct 2026.
- [Claude routines: what they are, triggers, cost and limits](https://munderdiffl.in/blog/claude-code-routines/): Claude routines run a saved Claude Code prompt in the cloud on a schedule, an API call or a GitHub event. Plans, steps and limits, checked 12 Oct 2026.
- [Claude Code Projects: what they are and who can use them](https://munderdiffl.in/blog/claude-code-projects/): Claude Code Projects, checked 11 Oct 2026: a public beta on Pro and Max where Claude runs your tasks as parallel threads. Access, limits and plan use.
- [Claude Managed Agents dynamic workflows: limits and cost](https://munderdiffl.in/blog/claude-managed-agents-dynamic-workflows/): Dynamic workflows let a Claude Managed Agents agent write a program that runs many agents. How to turn them on, beta limits and cost, checked 11 Oct 2026.
- [Claude Code agent teams: what they are, how to use them](https://munderdiffl.in/blog/claude-code-agent-teams/): Claude Code agent teams, checked 8 Oct 2026: what they are, the experimental setting that turns them on, how they differ from subagents, token cost and limits.
- [Claude Haiku 5.5: price, benchmarks and where it fits](https://munderdiffl.in/blog/claude-haiku-5-5/): Claude Haiku 5.5 is Anthropic's new small model. Its tiered price, Anthropic's own scores and how to pick it in Claude Code, checked 8 Oct 2026.
- [GPT-6 in ChatGPT: who gets it, when, and Intelligent UI](https://munderdiffl.in/blog/gpt-6-intelligent-ui/): GPT-6 with Intelligent UI in ChatGPT: Plus, Pro, Business and Enterprise from 7 Oct 2026, then Free and Go. Codex models are not changing in this release.
- [Claude Agent SDK: what it is, languages, cost and limits](https://munderdiffl.in/blog/what-is-claude-agent-sdk/): The Claude Agent SDK is Claude Code as a Python and TypeScript library. Install commands, features, billing, licence and limits, checked 8 Oct 2026.
- [ZCode: Z.ai's coding agent harness, explained](https://munderdiffl.in/blog/what-is-zcode/): ZCode is Z.ai's Apache-2.0 coding workspace: desktop app, browser view and terminal agent. How to install it, which models it lists, and what it costs.
- [OpenAI Decisions API: what it is, price and limits](https://munderdiffl.in/blog/openai-decisions-api/): The OpenAI Decisions API is a beta endpoint that returns a probability, a pick or a score. Price, limits and how to call it, checked 7 Oct 2026.
- [Mistral Large 4 (Le Chonk): size, price, scores and weights](https://munderdiffl.in/blog/mistral-large-4/): Mistral Large 4 is a 1 trillion parameter open weight model in public preview. Context window, API price, Mistral's own scores and weights, checked 6 Oct 2026.
- [OpenAI text watermarking: textGrain, who gets it, limits](https://munderdiffl.in/blog/openai-text-watermarking/): OpenAI text watermarking (textGrain) is opt in for the API now and is coming to eligible ChatGPT and Codex text in the EU. Limits checked 6 Oct 2026.
- [Reflection AI Beam: the 501B open weight model, explained](https://munderdiffl.in/blog/reflection-beam/): Reflection AI Beam is a 501B parameter open weight model. Size, Reflection's own scores, licence and when you can get it, checked 6 Oct 2026.
- [What Is Claude Cowork? Plans, Prices and How to Use It](https://munderdiffl.in/blog/what-is-claude-cowork/): Claude Cowork is Anthropic's agent for multi-step office work. Which plans include it, what it costs, where it runs and how to start, checked 3 Oct 2026.
- [Gemini Spark: what it is, what it costs and who can get it](https://munderdiffl.in/blog/what-is-gemini-spark/): Gemini Spark is Google's personal AI agent in the Gemini app. Plans, US prices, countries, limits and how it compares with Grok Bot, checked 3 Oct 2026.
- [Hermes Agent: what it is, how to install it and what it costs](https://munderdiffl.in/blog/what-is-hermes-agent/): Hermes Agent is Nous Research's MIT licensed AI agent with memory, skills and chat channels. Install command, models, pricing and how it compares.
- [What Is Codex? OpenAI's Coding Agent in the CLI, App and Cloud](https://munderdiffl.in/blog/what-is-codex/): What is Codex? OpenAI's coding agent runs in your terminal, the ChatGPT desktop app, your IDE and the cloud. Which plans include it, checked 2 Oct 2026.
- [What is Pi agent? The open source coding agent harness, explained](https://munderdiffl.in/blog/what-is-pi-agent/): Pi agent is Earendil's MIT licensed coding agent harness. What it is, how to install it, what Pi 1.0 and Pi Durable add, and how it compares.
- [What Is ChatGPT Dots? OpenAI's Always On Agents Explained](https://munderdiffl.in/blog/what-is-chatgpt-dots/): A ChatGPT dot is an always on agent with its own cloud computer. Needs Pro (from $100 a month) or Business Premium, not Plus. Not on Pro in the UK or EU.
- [What Is OpenCode? Install, Pricing and How It Compares](https://munderdiffl.in/blog/what-is-opencode/): OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
- [Is Ollama Good for Coding? Yes, for Scoped Work on the Right Model](https://munderdiffl.in/blog/is-ollama-good-for-coding/): Ollama is good for coding on scoped, private work if the model fits your memory. Which models to pull, how much RAM, and where Claude still wins.
- [Claude Max Plan: Is It Worth It? Max 5x vs 20x vs Pro](https://munderdiffl.in/blog/is-claude-code-max-worth-it/): Is the Claude Max plan worth it? Max 5x and 20x against Pro, checked 29 Sep 2026: price, five hour and weekly limits, Fable, and the API break even.
- [Is Claude Code Open Source?](https://munderdiffl.in/blog/is-claude-code-open-source/): No. Claude Code is proprietary under Anthropic's terms. What its GitHub repo holds, what the 2026 leak changed, and which open tools sit around it.
- [What Is Vibe Coding? Meaning, Risks and Safer Habits](https://munderdiffl.in/blog/what-is-vibe-coding/): Vibe coding means building software from AI prompts without reading the code. Where the term came from, when it is fine, where it bites, and what to check.
- [What Is an MCP Server?](https://munderdiffl.in/blog/what-is-an-mcp-server/): An MCP server gives an AI agent tools, data and prompts over one open protocol. What it is, the host, client and server roles, and where Claude Code plugs in.
- [Claude Code Pricing (Sep 2026): Every Plan, and Is It Free?](https://munderdiffl.in/blog/how-much-does-claude-code-cost/): Claude Code pricing checked 29 Sep 2026: Pro, Max, Team, Enterprise and API rates, usage limits, what changed this year, and whether Claude Code is free.
- [Loop Engineering: Designing Agent Loops That Converge](https://munderdiffl.in/blog/loop-engineering-for-ai-agents/): Everyone engineers the prompt; almost nobody engineers the loop around it. Stop conditions, drain loops, retry with backoff, compaction cycles, breaker escalation, budgets, and human gates — the outer-loop mechanics that make an agent converge instead of run away.
- [Review Agent Work Where It Happens](https://munderdiffl.in/blog/review-agent-work-where-it-happens/): AI agents produce diffs faster than humans can review them, and alt-tabbing to an external editor breaks supervision flow. The case for review-in-place: Munder Difflin's built-in Monaco IDE puts a CHANGES rail and side-by-side diffs vs HEAD right on the office floor.
- [Voice Is a Control Plane, Not a Gimmick](https://munderdiffl.in/blog/voice-as-a-control-plane-for-agent-fleets/): Voice is a terrible way to write code and a great way to run a fleet. Why low-bandwidth commands over high-bandwidth work is the right split — with Munder Difflin's Talk mode (echo-back confirmation, spend caps, michael-voice attribution) as the case study.
- [What Is an Agent Development Environment (ADE)?](https://munderdiffl.in/blog/what-is-an-agent-development-environment/): ADE means two things in 2026: a platform for building agents (Letta's sense) and a workspace for shipping code with fleets of coding agents (Orca's sense). Here's the full tooling map — chat IDEs, agent CLIs, agent IDEs, and agent harnesses — and how to pick.
- [What Is Harness Engineering?](https://munderdiffl.in/blog/what-is-harness-engineering/): Harness engineering is the discipline of building everything around the model — PTY plumbing, lifecycle hooks, mailboxes, memory, budgets, human gates, observability — because that's where agent reliability actually comes from. A definition, and four case studies from inside Munder Difflin.
- [Shareable Agent Roles: A Portable Format for Hiring AI Coworkers](https://munderdiffl.in/blog/shareable-agent-roles/): Why agent roles should be portable: what a hire manifest encodes, how one-click hiring stays safe, and how The Hiring Fair turns tacit setup into a shareable artifact.
- [The Clone Army Is a Trap: Why a Mixed Swarm Beats Ten Identical Agents](https://munderdiffl.in/blog/the-clone-army-trap-mixed-swarm-vs-identical-agents/): Spinning up tens of identical top-tier CLI agents feels powerful and burns tokens on work a cheaper agent could do. A mixed-capability swarm with shared memory wins on cost — and often on quality.
- [Why CLI Agents Are So Powerful — and How Munder Difflin Spends Fewer Tokens Doing the Same Work](https://munderdiffl.in/blog/why-cli-agents-are-powerful/): CLI agents are powerful because they have terminal-level access: they run builds, tests, and git, and verify their own work by executing it. Here's why that matters — and the concrete ways Munder Difflin cuts token consumption while doing it.
- [Generative Engine Optimization: The Techniques the Research Backs](https://munderdiffl.in/blog/generative-engine-optimization-techniques/): What the first peer-reviewed GEO study found works to get content cited by AI — statistics, source citations, quotes — and why keyword stuffing backfires.
- [The Agents Rule of Two: A Simple Safety Rule for Coding Agents](https://munderdiffl.in/blog/agents-rule-of-two/): Meta's Agents Rule of Two for coding agents: don't let one session combine untrusted input, private data, and external reach at once — or supervise.
- [Can Claude Code Agents Talk to Each Other?](https://munderdiffl.in/blog/can-claude-code-agents-talk-to-each-other/): Can Claude Code agents talk to each other? By default they report to their launcher — but a small coordination layer lets them message peer-to-peer.
- [Do More With Less: Smart Model Routing Across Agents](https://munderdiffl.in/blog/do-more-with-less-model-routing/): Why a hive of agents shouldn't all run the biggest model — and how routing the right task to the right model cuts cost and latency without losing quality.
- [How AI Agents Verify Their Own Work (Before Saying 'Done')](https://munderdiffl.in/blog/how-ai-agents-verify-their-own-work/): The discipline that makes an autonomous coding agent trustworthy: prove every claim, reproduce the green, and check the fix — not just the intent.
- [Long-Running Agents: The 2026 Shift from Minutes to Hours](https://munderdiffl.in/blog/long-running-agents-the-2026-shift/): Agents used to answer in seconds; now they run for hours. The data behind the shift — and why a longer agent is a different problem, not just a bigger one.
- [Building AI Dev Tools in the Open, On Purpose](https://munderdiffl.in/blog/open-source-ai-tools-on-purpose/): Agents touch your code, keys, and memory. That's why agent tooling should be open source and local-first — so you can verify it, not just trust it.
- [AEO for Dev Tools: Getting Cited by ChatGPT and Claude](https://munderdiffl.in/blog/what-is-aeo-for-dev-tools/): Answer Engine Optimization for dev tools: how to get cited by ChatGPT, Claude, and Perplexity with the right robots.txt, JSON-LD, and writing patterns.
- [When Do AI Agents Actually Need an Agent Protocol?](https://munderdiffl.in/blog/when-do-agents-need-an-agent-protocol/): When do AI agents need a protocol like MCP or A2A? Agents you own coordinate fine with file mailboxes; protocols earn their keep across boundaries.
- [The AI Coding Agent Glossary: Harness, Orchestrator, Hive, Subagent](https://munderdiffl.in/blog/ai-agent-glossary/): Plain-English definitions of the AI coding agent terms everyone trips over — harness, orchestrator, hive, subagent, agent memory — each in one line.
- [Munder Difflin FAQ: Everything People Ask](https://munderdiffl.in/blog/munder-difflin-faq/): Answers to the top Munder Difflin questions — what it is, is it free, does it run locally, which platforms, and how it differs from many terminals.
- [What Are Claude Code Agents? And How to Use Many at Once](https://munderdiffl.in/blog/what-are-claude-code-agents/): Claude Code agents explained in plain English — what an agent actually is, how subagents differ, and the leap from one agent to a coordinated team.
- [What Is an AI Agent Harness? Single vs Multi-Agent Harness Explained](https://munderdiffl.in/blog/what-is-a-multi-agent-harness/): An agent harness is the code around an AI model that runs its tool loop, memory and permissions. Single vs multi-agent harnesses, compared and dated.
### guides
- [How to Install and Use Munder Difflin: A Beginner's Guide](https://munderdiffl.in/blog/how-to-install-and-use-munder-difflin/): A plain language guide to Munder Difflin 0.5.2. What a coding agent is, which AI engine to pick based on what you already pay for, the one terminal command you'll ever run, and how to install and set it up on macOS, Windows or Linux.
- [Your First Hour With Munder Difflin: Onboarding to a Pro Office](https://munderdiffl.in/blog/your-first-hour-with-munder-difflin/): Munder Difflin 0.5.2 from download to a working Pro office: install Claude Code, Codex or OpenCode, sign in, go Pro, then automate reviews, Slack and email.
- [How to Get the Most Out of Your Claude Code Max Plan](https://munderdiffl.in/blog/claude-code-max-plan-tips/): Stretch a Claude Code Max plan: model choice, context hygiene, prompt caching, subagents and unattended runs, checked against Anthropic's own pages.
- [How to Make the Most of Your Codex Max Plan](https://munderdiffl.in/blog/codex-max-plan-tips/): There is no Codex Max plan. What ChatGPT Pro 5x and 20x give you in Codex, how the weekly limit works, and how to stretch it.
- [Gemini 4 Argon Pricing: Half Price Until 30 November, Then the Real Bill Arrives](https://munderdiffl.in/blog/gemini-4-argon-pricing/): Gemini 4 Argon pricing, checked 13 Oct 2026: $2 input and $10 output per 1M tokens through 30 Nov, then $4 and $20. No free tier. Worked examples inside.
- [How to Use Gemini 4 Argon: Five Ways In, and Free Is Not One of Them](https://munderdiffl.in/blog/how-to-use-gemini-4-argon/): How to use Gemini 4 Argon on 13 Oct 2026: the Gemini app, AI Mode, AI Studio, the API with model ID gemini-4-argon, and Antigravity. What each route needs.
- [ASD-STE100 Prompting: Karpathy's 4 Tricks for Clearer AI Answers](https://munderdiffl.in/blog/asd-ste100-prompting/): Ask your AI to answer in ASD-STE100, the aerospace manual standard, about 80% of the way. Plus Karpathy's three other tricks: diagrams, HTML and video.
- [Codex Pricing: Every Plan, Usage Limit and API Rate (Oct 2026)](https://munderdiffl.in/blog/codex-pricing/): Codex pricing checked 5 Oct 2026: which ChatGPT plans include Codex, US prices from $0 to $500, how usage limits are counted, credits and API key rates.
- [How to Use Claude Code: Setup, First Task and the Commands That Matter](https://munderdiffl.in/blog/how-to-use-claude-code/): How to use Claude Code, checked 5 Oct 2026: install it, sign in, run a first task, then the commands, shortcuts and habits a beginner needs.
- [Build Your Personal AGI on Your Computer Using Claude Code and Munder Difflin](https://munderdiffl.in/blog/build-your-personal-agi-on-your-computer/): Turn Claude Code into a personal AGI on your own computer: agents with memory, schedules, Gmail, Calendar, voice and Slack, set up in Munder Difflin
