Client handoff: giving Claude the context it needs to work on your accounts

Every agency has the same problem when they start using Claude: the output is competent but generic. It does not sound like the client, it avoids the client's actual constraints, and it occasionally suggests things the client has already rejected three times.

The fix is not better prompting in the moment. It is a structured client context document that you pass as the system prompt every time you work on that account.

What the client context document contains

Think of this as the briefing document you would give a new freelancer who needs to write for your client from day one — not a brand guidelines PDF, but the actual working knowledge your team has built over the engagement.

Voice and tone. Two or three sentences describing how the brand sounds. Not "professional and approachable" (useless) but "writes like a confident peer, not an authority figure — uses contractions, avoids jargon, occasionally challenges the reader." Then include three short examples: a social post, an email subject line, a CTA. Claude uses examples more than it uses descriptions.

Audience. Who is actually reading this. Not "busy professionals" but "VP-level buyers at mid-market SaaS companies who are skeptical of vendor content and read it on their phone during commute." The more specific, the better.

Constraints and rejections. Things the client has explicitly said no to. "Does not use the phrase 'game-changer.'" "Never leads with a question in subject lines." "Does not mention competitors by name." This is the most underused section — and the one that prevents the most wasted drafts.

Approved terminology. Words they use and words they avoid. Some clients have very specific language around their product category or audience. If they call customers "members" not "users," Claude needs to know.

What success looks like. The actual outcome the deliverable is meant to produce: "Email campaign goal is re-engagement, not awareness — we want a response or a click, not brand recall."

How to structure this as a system prompt

You are a copywriter working on behalf of [Client Name].

BRAND VOICE
[Two to three sentences. Include three examples.]

AUDIENCE
[Specific description of who reads this content.]

CONSTRAINTS
- Never use: [list]
- Always use: [list]
- Previously rejected angles: [list]

APPROVED TERMINOLOGY
[Key product/audience terms they use]

CURRENT TASK CONTEXT
[Fill this in per session: what campaign, what stage, what has been approved]

The "Current Task Context" section at the bottom is what changes each session. The rest stays stable and gets pasted at the top of every Claude conversation for this client.

Maintaining the document

The client context document only works if it is maintained. Three practices that keep it current:

Add rejections immediately. When a client rejects something for a recurring reason ("we don't want to sound urgent"), add that to the constraints section before you close the session. Two minutes now saves an hour next month.

Review it before pitches. Before a new campaign, read the constraints section. What did they push back on last time? What angle worked well? Update the document with what you learned.

Version it per campaign type. Some clients behave differently for organic social vs. paid ads vs. email. You may need a base document plus a campaign-specific overlay for each channel.

What this changes

The difference between a generic Claude prompt and a client-context Claude prompt is not subtle — it is the difference between a first draft that goes back to the client and a first draft that gets revised once and then goes. Agencies that build this system find that their Claude output quality compares favorably with a mid-level copywriter who has been on the account for three months. That is the standard to aim for.

Further reading

• Collaborate with Claude on Projects — using Projects to manage client context