The prompt
Paste this into Claude, ChatGPT, Cursor, or any MCP-compatible client connected to your Bourd account. Replace the highlighted placeholders with your own values before sending.
I'm onboarding a new client into the {{ agency account }} account on Bourd. Use this conversation to stand up their workspace, register their competitors, and seed a starter prompt set.
Brief:
- Client: {{ client name }} ({{ client website }})
- Category: {{ category description }}
- Buyer: {{ buyer description }}
- Competitors: {{ competitor 1 }} ({{ competitor 1 website }}), {{ competitor 2 }} ({{ competitor 2 website }}), {{ competitor 3 }} ({{ competitor 3 website }})
Before creating anything in Bourd, work out what the prompts should be. Read the client's site and each competitor's site. Get a feel for how each company positions itself, who the buyer actually is, and the questions that buyer would ask in ChatGPT or Claude. Buyers ask in chat using full sentences with context, often naming their role, segment, or specific situation directly.
Propose around 20 prompts grouped into three buckets:
- 6 to 8 branded prompts that mention {{ client name }} by name
- 8 to 10 category-entry prompts a buyer would ask before they know {{ client name }} exists
- 4 to 6 comparison prompts that name one or more of the registered competitors
Show me the proposed set. We can refine wording, swap prompts in or out, and adjust the mix per bucket before anything gets created.
Once I sign off, create the workspace called "{{ client name }}", register the competitors, create three tags ("branded", "category-entry", "comparison"), and write the prompts with the matching tag applied to each. Don't run any prompts yet. 14 placeholders to fill in. The agent will call: select_account, create_workspace, create_prompt, and 3 more.
What this workflow does
The agent reads the brief and the listed sites, then proposes around 20 prompts grouped into branded, category-entry, and comparison buckets. You refine the wording and the mix until it's right. On your sign-off, the agent creates the workspace, registers the competitors, creates a tag for each bucket, and writes the prompts with the matching tag applied. The set lands in your dashboard already filterable by intent. Nothing runs until you start the prompts yourself.
Why it works
Workspace setup is the same shape every time: a fixed sequence of create calls with predictable inputs. Strategists click through quickly because none of the steps feel important on their own, which is when tagging gets skipped and competitor lists go in incomplete. The agent runs the chain at machine reliability and stops where review actually pays off.
What a good 20-prompt seed looks like
A working mix:
- Six to eight branded prompts that mention the client by name.
- Eight to ten category-entry prompts a buyer would ask before they know the client exists.
- Four to six comparison prompts that name the registered competitors.
Skip vague prompts like "best CRM" that no real buyer types into Claude. Push the agent to write in the phrasing a buyer would actually use, with the segment, role, and use case named explicitly.
Common failure modes
Three patterns to watch for:
- Generic prompts. "What is content marketing" tracks nothing useful. Ask the agent for prompts that include the category, the buyer role, and a specific use case.
- Invented competitors. Pass the competitor list verbatim. A "top three competitors" instruction lets the agent invent a plausible-sounding fourth name.
- Reused prompt sets. Each new workspace needs a fresh brief: the client's target customer, geography if relevant, and the slice of the category they actually compete in.
How to extend it
- Save a brief template per segment in your handoff doc. Swap in the client name, competitor list, and any segment-specific framing on each run.
- After approval, run a follow-up turn: "list the prompts grouped by intent and flag any that overlap or that I should delete."
- Chain into Slack or Notion so the prompt set posts to a strategist channel for a second pair of eyes before any prompts run.
Tools the agent calls
The agent picks from these 6 Bourd MCP tools based on the prompt. You do not call them directly.
select_account create_workspace create_prompt create_competitor create_tag tag_prompts
Pairs with
MCP agents chain across servers. Add these alongside Bourd and the same prompt can hand output straight to the tools you already work in.
Slack Notion Google Docs Linear