Things to ask your assistant
Practical prompts for reading results, preparing campaigns and finding leads.
Connect your client first. Replace the bracketed values below with your workspace or campaign ID. These prompts describe your intent; your signed-in access level or static key scopes enforce access.
Create a CRM follow-up journey
In workspace [workspace ID], create a Person and Company from the details I provide, read them back, link them, and create a Deal in the pipeline and stage I approve. Link both records to the Deal, then add a Note and an open follow-up Task. Show me each proposed mutation before running it. Use a new stable idempotency key for every logical mutation and preserve each key for an identical retry. Finally, read everything back and report the IDs. Do not delete, merge, message or call anyone.
This requires Full access on a signed-in connection, or a workspace key with read/write scopes for People, Companies, Deals, Notes and Tasks plus read scopes for pipelines, members and custom fields. Complete the Task later with update_task and body.status: completed. Move the Deal only after reading its current stage and sending that value as expectedCurrentStageId.
Check results without changing anything
In workspace [workspace ID], list my campaigns. For campaign [campaign ID], show its results and explain the metrics. Do not make any changes.
Uses list_campaigns and get_campaign_metrics. Understand campaign results.
Create a draft without launching it
In workspace [workspace ID], create a draft called September interviews. Ask me for any required details that are missing. Leave it as a draft and do not search for leads or launch it.
Requires Full access, or a workspace key with campaigns:write. Creating the draft does not send a campaign, but configured campaign-created automations may run.
Prepare and review a sequence
For draft campaign [campaign ID] in workspace [workspace ID], show me the email steps you propose before attaching them. Do not launch the campaign.
The assistant can use set_campaign_sequence for a draft without an existing sequence. Review the saved sequence in Acquisity; the API does not return its contents. Sequence limitations.
Preview leads before spending credits
In workspace [workspace ID], propose filters for UK marketing agency owners at companies with 10–50 employees. Show the filters for review, then preview the match count. Do not run a paid search or attach any leads.
Uses ai_search_leads and preview_lead_search, with Read-only or Full access. For a static-key connection, use a workspace key: standalone tools require leads:read; campaign-specific previews require campaigns:write.
Run an approved paid search
Use the filters I approved in workspace [workspace ID] to search for at most 100 leads and export them without adding them to a campaign. This paid search is approved. Use one idempotency key, keep it for any retry, and give me the operation ID and download link when complete.
Uses search_leads with leads:search. Only use this prompt when you intend to spend credits. Wait for the operation to succeed; use get_operation if it is still running.
Import a personalized CSV into a draft
Read prospects.csv and use the workspace I choose from
list_workspaces. Create a draft campaign named Personalization demo. Map Email to email, First name to firstName, Company to company, Opening line to customVariables.icebreaker, and Reason to customVariables.reasonToReachOut. Show the mapping and flag missing values or unmapped columns. Upload the mapped objects withadd_campaign_leadsin supported batches, each with a stable retry key. Poll every returned operation, then read back two leads to verify customVariables. Attach one email step that uses{{firstName}},{{companyName}},{{icebreaker}}, and{{reasonToReachOut}}. Validate the campaign. Leave it as a draft and do not launch or send.
This requires Full access. For a static-key connection, it requires a workspace key with workspaces:read, campaigns:read, campaigns:write, and campaignLeads:read. The client must be able to read the CSV: Claude Code can use authorized local file tools, while Claude Desktop needs an accessible attachment or supplied lead objects. CSV mapping, limits, and retries.
Check readiness before launch
Validate campaign [campaign ID] in workspace [workspace ID]. Explain every issue and suggest what I should fix. Do not launch it.
Uses validate_campaign with campaigns:read. A ready result does not guarantee that every provider-side check will pass during launch.
When you explicitly want to send:
Launch campaign [campaign ID] in workspace [workspace ID]. I approve sending real outreach. Use a stable idempotency key and return the operation ID and status.
This requires Full access, or campaigns:write on a static key, and can send real email and spend money. Launch and pause guide.
Pause a campaign
Pause campaign [campaign ID] in workspace [workspace ID]. Report the operation status and tell me if anything still needs attention.
Uses pause_campaign. Messages already handed off to the sending provider may still be delivered.
Check work that is still running
Check operation [operation ID]. If it is still running, report that status without submitting the original change again. If it failed, explain the error and suggested next step.
Uses get_operation. For campaign lead searches, a succeeded operation means the search was dispatched, so inspect the campaign's leads to see when attachment has finished.