@heynavtoor: https://x.com/heynavtoor/status/2053422550567502046
Summary
This article introduces a framework for using seven specific Claude sub-agents to automate roles such as research, editing, project management, and financial analysis, effectively replacing a high-cost team.
View Cached Full Text
Cached at: 05/10/26, 02:29 PM
The 7 Claude Sub-Agents That Replace a $200K Team
You do not need to hire. You need to delegate to seven files.
Save this. You will read it twice.
A sub-agent is a markdown file that gives Claude a job, a brain, and a set of rules. You drop it in a folder. Claude can call it on demand. Each sub-agent runs in its own context window with its own instructions, so it does not contaminate your main thread and does not forget who it is.
Most people use Claude as one assistant. The smarter move is to use Claude as a team of seven.
Below: 7 sub-agents. Each one replaces a real role with a real salary. Each one comes with the complete agent file. Copy. Save. Ship.
How to Install Any Sub-Agent in This List
Three ways. Pick yours.
Claude Code: Create a folder at .claude/agents/ in your project. Save each agent as <name>.md. Claude finds them automatically. Call one with /agents or let Claude pick the right one.
Claude.ai: Settings. Sub-Agents. Add. Paste the file contents.
Claude Desktop and Cowork: Same as Claude.ai. Settings. Sub-Agents. Add.
Build once. Run forever.
The 7 Sub-Agents
Total payroll replaced: $780,000 a year. The title says $200K because that is what one founder, freelancer, or operator can realistically save right now. The full $780K is what it costs a real company to hire all seven.
You are about to run that company from a folder.
01. The Researcher
Replaces: A $90K Research Analyst.
What it does: Goes deep on any topic. Pulls primary sources. Returns a structured brief with three findings, three contradictions, and three open questions.
Fire it when: You need a real briefing before a meeting, a pitch, an article, or a decision. Not a summary. A briefing.
02. The Editor
Replaces: An $85K Senior Editor.
What it does: Reads your draft like a hostile editor. Cuts filler. Tightens hooks. Flags weak claims. Returns the same piece, 30% shorter, twice as sharp.
Fire it when: You finished a draft and you know it is bloated but you are too close to see it. Stop publishing first drafts.
03. The Project Manager
Replaces: A $110K Project Manager.
What it does: Takes a goal. Returns a week-by-week plan with owners, deadlines, and the one thing that kills the project if it slips.
Fire it when: You are about to start something that will take more than a week and involve more than one person. Plan once. Ship clean.
04. The Analyst
Replaces: A $120K Data Analyst.
What it does: Point it at a CSV, a dashboard paste, or a wall of numbers. Returns the story, the outliers, the so-what, and the chart you should actually build.
Fire it when: You exported a report and you are staring at it without knowing what to say in the meeting. Get the headline first. Build the chart second.
05. The Recruiter
Replaces: A $95K Talent Sourcer.
What it does: Give it a role. Returns a sourcing plan, the channels, the outreach template, the screening rubric, and the rejection email so you stop ghosting.
Fire it when: You posted the role and got 200 bad applicants and zero good ones. The sourcing was wrong, not the role.
06. The Ops Lead
Replaces: A $100K Operations Manager.
What it does: Audits any process you describe. Finds the three steps to automate, the two to kill, and the one to never touch. Returns SOPs, not vibes.
Fire it when: A process feels heavy and you cannot say why. Audit before you build. Building on top of a bad process just gives you a faster bad process.
07. The CFO
Replaces: A $180K Fractional CFO.
What it does: Reads your numbers. Returns runway, burn, the line item bleeding you dry, and what to cut first. No spreadsheet theater. The call.
Fire it when: The bank balance is making you nervous and the spreadsheet has too many tabs. Get the call first. Build the model second.
Claude is the engine. Skills are the steering wheel. Sub-agents are the team in the back seat doing the work while you drive. Without them, you are a one-person company pretending to be a real one.
Which Three to Install First
Do not install all seven. You will not use them. Pick three.
Solo founder: CFO, Project Manager, Recruiter. Money, plan, people. The trinity.
Freelancer or consultant: Researcher, Editor, Project Manager. Sell the brief, ship the work, hit the deadline.
Engineer or builder: Researcher, Analyst, Ops Lead. Read the field, read the data, fix the workflow.
Content creator: Researcher, Editor, Analyst. Find the story, sharpen it, measure what worked.
Operator inside a company: Project Manager, Ops Lead, Analyst. Run the project, fix the process, prove the impact.
Install three. Run them daily for two weeks. Add a fourth only when you catch yourself doing the same job twice. That is how you build a team without hiring one.
How a Day Actually Runs
Monday morning. You open Claude.
You ask Researcher for a brief on the company you are pitching Friday. It returns a one-pager with three findings and three contradictions. You read it on the train.
You ask Editor to tear apart the proposal you wrote Sunday night. It returns the same proposal, 30% shorter, with the opening rewritten. You ship.
You ask Project Manager to plan the launch you committed to last week. It returns the milestones, the owners, the critical path. You forward it to the team.
You ask Analyst what last week’s numbers actually say. It returns the headline, the outliers, and the chart. You drop the chart in the deck.
You ask CFO if you can afford the new hire. It returns the math, the runway impact, and the cut list. You decide.
That was Monday. You did the work of five people in 90 minutes. The week is yours.
This is what the docs do not tell you. The compounding is not in any single sub-agent. The compounding is in running them in sequence, every day, on the same business.
The Uncomfortable Truth
Most readers will save this and forget it.
A few will copy three sub-agents and use them tomorrow.
A smaller few will run all seven for 30 days and stop pretending they need to hire that next role.
The gap between those three groups is not money. It is not talent. It is whether you opened the folder and pasted the file.
The cost of inaction is the salary of the person you are about to hire to do work seven markdown files can already do.
Open the folder.
Follow for the Next One
If this gave you something the docs do not, follow @heynavtoor.
One of these every Tuesday and Friday. Skills, sub-agents, commands, the workflows that make Claude actually run your business.
Bookmark this. You will come back when you need to install Sub-Agent 4.
Repost it to the founder still trying to hire their way out of a workflow problem. You just gave them six months of runway.
No newsletter. No course. No funnel. Just the next one, in your feed.
TL;DR
7 sub-agents. $780K of replaced payroll. One folder.
Researcher. Editor. Project Manager. Analyst. Recruiter. Ops Lead. CFO.
The team is the file. Open the folder.
Similar Articles
@heynavtoor: https://x.com/heynavtoor/status/2054149098513866941
This article presents nine Claude agents that run overnight to handle tasks like briefings, research, and inbox triage, allowing users to wake up to completed work. It provides instructions for setting up each agent using Claude Code, Claude.ai, or Claude Desktop.
I built a multi-agent product team inside Claude — CEO, CPO, CTO, Senior Devs, QA, Code Reviewer all chained together
The author describes a multi-agent system built within Claude that simulates a full product team (CEO, CPO, CTO, devs, QA) to streamline software development and decision-making. The setup uses role-specific skills and strict validation to reduce rework and is packaged for easy installation.
@ashpreetbedi: https://x.com/ashpreetbedi/status/2053885390717890757
Ashpreet Bedi shares a new agent platform designed for auto-improvement, where coding agents like Claude Code can autonomously create, test, and refine other agents using a unified data and API stack.
@ashwingop: https://x.com/ashwingop/status/2052777467732283817
An analysis of Claude's Managed Agents as a harbinger for the next AI infrastructure layer—'Company Brain'—an operational state layer that enables agents and apps to act from shared company context, contrasting with simpler knowledge bases or markdown-based prototypes.
@corbin_braun: 7 AI Agents Build Entire Software
Demonstrates a system of 7 AI agents collaborating to build an entire software application.