Three Foundational Context Files Every Claude Cowork User Should Create

How to build persistent, compounding context for Claude by creating three plain-text files that define who you are, how you communicate, and how you want Claude to behave.

One of the most effective ways to get consistent, high-quality output from Claude in a cowork environment is to give it context that persists across every session. Rather than re-explaining yourself at the start of each conversation, you can create a small set of plain-text markdown files that Claude reads automatically when it starts. These files are simple to write, quick to update, and compound in value over time. Three of them cover most of what Claude needs to serve you well: who you are, how you communicate, and how you want Claude to behave.

  • What the three foundational context files are and what each one does
  • What to include in your professional biography file
  • How to describe your communication style and brand voice
  • How to specify your working style preferences and default behaviors
  • Where to store these files and how they build value over time

The following sections explain each file in detail and walk through the kinds of information that make them most useful.

This lesson is a preview from our AI for the Workplace with Claude Course Online. Enroll in a course for detailed lessons, live instructor support, and project-based training.

The aboutme.md file is your professional biography written for a colleague who needs to understand your world. It tells Claude who you are, what your role is, what organization you work for, what you are responsible for, what success looks like in your position, the kind of work you produce, and who your audience is.

Think of this file as the answer to the question Claude would ask if it could ask one thing before every session: who am I working with and what do they do? The more specifically you answer that question in this file, the better Claude can tailor its responses to your actual context rather than making generic assumptions.

The Second File: thebrandvoice.md

The thebrandvoice.md file tells Claude how you communicate. It captures your writing tone, whether that is formal, conversational, technical, or plain language. It documents the phrases you consistently use or consistently avoid. It describes your sentence style: short and punchy, or detailed and analytical. It notes any terminology that is specific to your industry or organization.

When this file is in place, Claude does not default to a generic writing style. It matches yours. If you write in short, direct sentences, Claude will write that way. If you prefer thorough, carefully structured prose, Claude will produce that instead. The file essentially gives Claude a stylistic blueprint so that its output sounds like it came from your voice, not from a default AI template.

The Third File: theworkingstyle.md

The theworkingstyle.md file tells Claude how you want it to behave during a session. This is where you specify the procedural preferences that apply to every interaction. For example: should Claude ask you clarifying questions before starting a task, or jump straight to its best attempt? What file formats do you prefer for outputs? How long should responses be by default? Should Claude use bullet points or prose? When Claude is uncertain about something, should it ask you, or should it make its best guess and flag the uncertainty?

These are the kinds of preferences that would otherwise need to be restated in every conversation. By documenting them in a single file, you make them persistent and automatic.

Where to Store Them and How They Compound Over Time

All three files can be stored in a dedicated folder within your main working directory. A folder named something like "Claude context" works well. Claude reads these files at the beginning of every session, which means any updates you make to them take effect immediately in subsequent conversations.

The real value of these files is that they compound over time. As you refine them, adding a line here, adjusting a preference there, correcting something that does not quite reflect how you work, Claude gets incrementally better at serving you with each update. What starts as a rough professional biography becomes a precise, accurate description of your role and context. What starts as a general style description becomes a detailed voice guide. The files get more useful the longer you work with them and the more carefully you maintain them.

  • Three plain-text markdown files give Claude persistent context across every cowork session: aboutme.md, thebrandvoice.md, and theworkingstyle.md.
  • aboutme.md tells Claude who you are, what your role is, and what success looks like in your position.
  • thebrandvoice.md tells Claude how you write, including tone, style, preferred phrasing, and words or phrases to avoid.
  • theworkingstyle.md tells Claude how to behave: whether to ask questions before starting, preferred output formats, default length, and what to do when uncertain.
  • Storing these files in a dedicated folder ensures Claude reads them at the start of every session.
  • These files compound in value over time. Each refinement makes Claude more accurately calibrated to your specific needs and working style.
photo of Brian Simms

Brian Simms

Brian Simms teaches for Graduate School USA in the area of Artificial Intelligence, helping federal agencies build the knowledge and skills needed to adopt AI responsibly and effectively. An AI educator and author, he focuses on practical, mission-driven applications of AI for government leaders, program managers, and technical professionals.

More articles by Brian Simms

How to Learn AI

Build practical, career-focused skills in AI through hands-on training designed for beginners and professionals alike. Learn fundamental tools and workflows that prepare you for real-world projects or industry certification.