The Freelancer's CLAUDE.md: A Complete Template for Client Work

Every CLAUDE.md guide you can find online is written for software engineers running projects. They cover how to describe your tech stack, how to explain your test suite, how to instruct Claude on your preferred coding patterns. These are useful. They are also completely wrong for freelancers.

Freelancers do not run projects with a tech stack. They run 10 to 20 simultaneous client relationships, each with its own history, expectations, communication preferences, and outstanding commitments. Their CLAUDE.md does not need to explain npm run test. It needs to make Claude a genuinely useful thinking partner across those relationships.

This is the guide that has not been written. A CLAUDE.md architecture specifically for freelancers — with a full, copy-paste template at the end.

Why CLAUDE.md Changes Everything for Freelancers

Claude Code reads CLAUDE.md from your project directory before every session. It is the first thing Claude knows about you. Everything that comes after — every question you ask, every draft you request, every decision you explore — runs on top of this foundation.

Most freelancers do not have a CLAUDE.md. They paste in background when it feels relevant, which means the background varies by session, is often incomplete, and disappears the moment the conversation ends.

Every Claude session starts from zero. Without a CLAUDE.md, Claude does not know you are a freelancer. It does not know what you specialize in, who your clients are, how you work, what constraints shape your decisions, or what you are optimizing for right now. So it gives advice calibrated to a generic professional. Generic context produces generic answers.

A well-constructed CLAUDE.md changes that. It means every session starts with full context on who you are and how you operate. It is the difference between consulting a well-briefed advisor and consulting a stranger.

The second function of CLAUDE.md for freelancers is habit. Writing your CLAUDE.md forces you to be explicit about things that usually live in your head: your working principles, your constraints, your client tier logic, your decision-making framework. Most freelancers have never written these down. The act of writing them is clarifying — separate from the AI benefit.

What Goes in a Freelancer's CLAUDE.md

A freelancer's CLAUDE.md has four distinct layers. Each one answers a different question Claude needs before it can give useful advice.

• Layer 1: Identity — Who are you? What do you do? What does your business look like? Who do you serve?
• Layer 2: Client context architecture — Who are your clients? How are they tiered? What do you know about each one? Where does that knowledge live?
• Layer 3: Working style — How do you make decisions? What are your constraints? What are your communication principles? What should Claude never suggest?
• Layer 4: Current state — What are you focused on right now? What are the active pressures? What is the 30-day context?

Most guides only address the first layer. Layers 2 through 4 are where the leverage is for freelancers.

The Identity Layer

The identity layer tells Claude who it is working with. Not a vague summary — a specific description that shapes every subsequent interaction.

What belongs here:

• Your professional category and specialization. Not just "I'm a freelancer." What kind? What do you specialize in? What industries?
• The shape of your business. How many clients do you work with at once? What types of engagements — project-based, retainer, hybrid?
• Who you serve specifically. The profile of the clients you take. What they are trying to accomplish.
• What you are optimizing for right now. Revenue? Relationships? Specific skills? Transitioning between niches?
• What you are not. The work you do not take. The clients you say no to.

The forcing function for this section: If a recruiter read only your identity layer and had to describe you to someone else in 30 seconds, would they get it right? If the description is vague enough that it could apply to 500 other freelancers, it is not specific enough to be useful.

The Client Context Architecture

This is the layer that most freelancers have not thought about — and the one that produces the largest immediate improvement in Claude's usefulness.

The client context architecture does two things. First, it establishes a tiering system so Claude understands which clients matter most. Second, it tells Claude where detailed client information lives and how to find it.

Client tiering is essential because not all clients are equivalent. Tier 1 clients — your core long-term relationships — warrant more proactive attention, more relationship maintenance, and more careful communication. Tier 3 clients — occasional or lower-stakes work — have different rules. Claude should know this so it can calibrate advice accordingly.

Context file references are the bridge between your CLAUDE.md and detailed client knowledge. Instead of putting all client information into CLAUDE.md (which would be unwieldy and quickly stale), you create a reference system — individual files per client — and tell Claude where they live and how to use them.

The structure that works: one client file per active client. The file covers how you know them, the current project or engagement, recent key interactions, outstanding commitments, relationship notes, and their communication preferences. Store these somewhere consistent — ~/.claude/clients/ is clean and persistent.

When you ask Claude to help you draft a follow-up or prepare for a meeting, you say: "Read the client file for [Name] and help me prepare." The file does not need to be in your CLAUDE.md — it just needs to exist, and Claude needs to know where to find it.

Update cadence should also be in your CLAUDE.md. Not because Claude will enforce it, but because making the commitment explicit increases the likelihood you actually do it. Relationship context that is three months stale is worse than no context — it points Claude's attention at outdated information.

The Working Style Layer

The working style layer is where you make your decision-making principles, constraints, and communication philosophy explicit.

This is the layer most freelancers find uncomfortable to write because it requires being honest about how they actually operate — not how they think they operate.

Decision-making principles: How do you evaluate whether to take a new client? What is your rate policy — do you negotiate? Under what conditions do you accept lower-rate work? How do you decide when to end an engagement?

Constraints that shape your work: Do you work specific hours? Do you have non-negotiable capacity limits? Are there industries or project types you decline?

Communication defaults: What is your standard response time? How do you handle scope creep? What is your approach to difficult conversations with clients?

What Claude should never suggest: This is underrated. "Do not suggest I take on a fourth concurrent client." "Do not suggest I discount my rates." "Do not suggest I respond to client emails during weekends." These instructions save the tedium of pushing back on generic advice that does not fit your situation.

The working style layer makes Claude's suggestions actionable instead of theoretically correct. Generic advice is often advice that does not account for your constraints. Explicit constraints make Claude useful instead of academic.

The Current State Layer

The current state layer is the most frequently updated — and the most frequently omitted.

It answers: what is happening right now? What are you focused on this month? What are the live pressures and active decisions?

What belongs here:

• Current client load and capacity situation
• Live projects that are high-stakes or time-sensitive
• Active business decisions you are working through
• Open questions you keep coming back to
• The 30-day priority list

This layer should be reviewed monthly and updated whenever your situation materially changes. A CLAUDE.md with a current state section from six months ago is misleading — it tells Claude you are in a situation you are no longer in.

The forcing function: before any significant Claude session, spend 90 seconds checking whether the current state section still reflects reality. If it does not, update it. The session that follows will be materially better.

A Complete Template

This is a full, copy-paste CLAUDE.md template designed for freelancers. Replace the bracketed placeholders with your actual information. Remove sections that do not apply. Add sections specific to your work.

# CLAUDE.md — [Your Name]

Last updated: [Date]

---

## Who I Am

I am a [your specialty] freelancer specializing in [specific niche or domain]. I have been
doing this work for [X] years.

My business model: I work with [number] concurrent clients on [retainer / project / hybrid]
engagements. My clients are typically [client profile — industry, size, type of person you
work with].

I serve clients who are trying to [core outcome your clients want]. I am the right person for
this work when [conditions under which you add the most value].

**What I do not do:** [List 2-3 types of work or client situations you decline.]

**What I am optimizing for right now:** [Current business goal — e.g., "building two more
long-term retainer clients," "transitioning from X niche to Y niche," "maintaining current
revenue while reducing hours."]

---

## My Clients

### Tier 1 — Core Long-Term Relationships

These are clients I would protect at significant cost. They are long-term, high-trust
relationships that I actively invest in.

- [Client Name / Pseudonym] — [1-line description: type of work, length of relationship]
- [Client Name / Pseudonym] — [1-line description]
- [Client Name / Pseudonym] — [1-line description]

**Detailed context files for Tier 1 clients live at:** `~/.claude/clients/`
To access: read the relevant client file before any task involving that person.

### Tier 2 — Active Project Clients

Current clients with defined projects or shorter-term engagements.

- [Client Name] — [Project description and current status]
- [Client Name] — [Project description and current status]

### Tier 3 — Occasional or Prospective

Lower-frequency relationships — past clients who may return, warm leads, occasional work.
No dedicated files — context lives in my notes.

### Client File Update Cadence

After every significant interaction with a Tier 1 or Tier 2 client, I update the relevant
client file. Significant = any meeting, major email exchange, scope change, or commitment.

---

## How I Work

### Rate and Engagement Policy

- My standard rate is $[X]/hour or $[Y] for [scope].
- I do not negotiate on rate below $[floor].
- I will adjust scope, not price, when budget is a constraint.
- I decline work under $[minimum project size].

### Capacity and Availability

- My maximum concurrent client load is [number] active clients.
- I am available [days / hours].
- I do not respond to client communication outside [your boundaries].
- Current capacity status: [at capacity / one slot open / actively seeking / etc.]

### How I Make Client Decisions

When evaluating whether to take a new client or project, I weigh:
1. [Decision factor 1 — e.g., "fit with my specialty"]
2. [Decision factor 2 — e.g., "client communication style"]
3. [Decision factor 3 — e.g., "budget and scope clarity"]
4. [Decision factor 4 — e.g., "referral relationship"]

A new client is the right fit when [your positive criteria].
A new client is a red flag when [your warning signs].

### Scope and Delivery

- I define scope in writing before any project begins.
- Scope creep is acknowledged and repriced, not absorbed.
- [Any other delivery principles specific to your work.]

### Communication Defaults

- Standard client response time: [X hours/days].
- For complex client issues, I respond with a call, not a long email.
- When delivering bad news to a client, I [your approach].
- When a client relationship is not working, I [your approach to difficult conversations].

---

## Constraints Claude Should Know

**Do not suggest I:**
- Take on more than [number] concurrent clients
- Discount my rate as a competitive strategy
- Take project-based work when my capacity is full
- Respond to client communication outside my stated availability
- [Any other constraint specific to your situation]

**I have already decided:**
- [Any standing decisions that should not be re-litigated]

**My non-negotiables:**
- [1-2 hard limits that apply regardless of context]

---

## Current State — [Month, Year]

### Current Focus

[1-2 sentences on what you are most focused on right now.]

### Active High-Stakes Items

- [Client or project name]: [What is live, why it matters, what you are watching]
- [Client or project name]: [What is live, why it matters, what you are watching]

### Open Business Questions

- [Question you are actively working through]
- [Question you are actively working through]

### 30-Day Priority

1. [Most important thing to accomplish this month]
2. [Second priority]
3. [Third priority]

---

## What Claude Can Do For Me

### Client Communication

When drafting emails, proposals, or follow-ups:
- Read the relevant client file first
- Match [your communication style]
- Default to shorter over longer
- Do not hedge or soften delivery of factual information

### Decision Support

When I am working through a significant decision:
- Reference my client context and current state first
- Ask clarifying questions before suggesting options
- Present the trade-offs directly — I do not need options softened
- Be honest when my stated priorities conflict with what I am considering

### Strategy and Direction

When I ask for strategic input on my business:
- Reference what I have said I am optimizing for
- Do not suggest actions that violate my stated constraints
- Challenge me on inconsistencies between my priorities and my decisions

---

## Working Notes

[This section is for ad-hoc context that does not fit elsewhere. Temporary. Updated
frequently.]

Current notes:
- [Add notes here as needed]

What CLAUDE.md Cannot Do

A CLAUDE.md is a static document. It cannot update itself. It cannot capture what happened in the meeting you had yesterday. It cannot synthesize your Gmail history with a client. It cannot surface the fact that Tier 1 client Chen has gone quiet for three weeks.

The coverage problem is real. Your CLAUDE.md covers the relationships you explicitly wrote down. The email thread where you made an offhand commitment — not in the system. The calendar event linked to a client project — not in the system. The tone shift in Marcus's last three emails that might signal something — not in the system.

A CLAUDE.md is the foundation. It is not the full architecture.

The staleness problem is also real. A CLAUDE.md written in January describes your January. By March, your biggest client may have churned and your priority project may have shipped. Review the full document monthly. Update the current state section whenever something material changes.

The Next Layer

A well-maintained CLAUDE.md is meaningfully better than nothing. It gives Claude a real foundation. It produces sessions that start from context instead of zero. It makes the advice more specific, the drafts more calibrated, the decisions better-informed.

The ceiling of manual CLAUDE.md management shows up around six active clients. At that point, you are maintaining individual client files for six or more people, updating them consistently after every significant interaction, and keeping your CLAUDE.md current state section accurate. That is viable. It is also the full-time maintenance job you hired an AI to replace.

The automated alternative is a system that builds and maintains this context for you. Gmail sync that links your email threads to your client records. Calendar sync that connects meetings to client files. Conversation intelligence that analyzes transcripts, extracts commitments, and updates relationship records automatically. A decision log that tracks what you decided and what actually happened.

That is what Software of You does. It is a Claude Code plugin — running in the same environment as your CLAUDE.md — that automates the context engineering your CLAUDE.md starts. $149 one-time. Local SQLite database. No subscription, no lock-in, no cloud.

The CLAUDE.md you build today is the identity layer. Software of You is the full system.