search-reports

<!-- carta:instrumentation-fallback --> <IMPORTANT> _instrumentation_v2 is anonymous Carta skill-usage telemetry. ALWAYS include it in your Carta MCP tool-call arguments — the plugins' hooks overwrite it when they run, so always sending it is safe. Shape: _instrumentation_v2: {"skills": ["<plugin>:<skill>", ...], "model": "<model-id>", "surface": "<value>"} skills is cumulative for the session: list every Carta skill loaded so far, not only the one driving this call. Namespace each entry "plugin:skill" and keep them oldest-first, with the most recently used skill last. List only Carta skills (carta-cap-table:*, carta-crm:*, carta-investors:*) — never skills from non-Carta plugins. surface is the Claude surface you are running in: "chat" (claude.ai or the Claude app, i.e. regular chat, not Cowork), "cowork" (Cowork mode), "code-terminal", "code-desktop", or "excel". Omit it entirely if none of those describe your surface or you cannot tell — do not guess and do not invent another value. </IMPORTANT>

Overview

A saved report is a stored filter plus a stored column set over one entity: deals, contacts, companies, investors, fundraisings or fees. Running one renders its rows as a table, using the columns the report saved.

Reports are not searched by keyword. They are listed, chosen, then run — so Step 1 always precedes Step 2.

For an aggregated breakdown of deal flow by sector, source or owner rather than the rows themselves, use the deal-flow-analytics skill instead.

Step 1 — Find the report

Read the list through crm_call_tool, never crm_view_tool. list_reports has no view of its own, and this step is for you rather than the user: you need each report's id and entityType to run it.

crm_call_tool({ "name": "crm:list_reports", "arguments": {} })

Pass entityType to narrow the list when the user named one — "deal", "contact", "company", "investor", "fundraising" or "fees":

crm_call_tool({ "name": "crm:list_reports", "arguments": { entityType: "deal" } })

Then branch on what came back:

• One clear match on name → Step 2.
• Several plausible matches → do NOT guess. Name them and ask which one. Running the wrong report answers a question nobody asked, and looks authoritative doing it.
• No match → say so, and say what the reports are actually called. Never invent a report or a report id.
• The user only asked what reports exist → answer from this list and stop. There is nothing to render.

Step 2 — Run the report

Use the id and the entityType from the entry you matched in Step 1. Both are on that entry; never guess either.

crm_view_tool({
  "name": "crm:get_report_data",
  "arguments": {
    reportId: "<id from list_reports>",
    entityType: "<entityType from the same entry>",
    limit: 50
  }
})

The report's own saved columns drive the table. Do not pass a column list, and do not ask the user which columns they want — the report already answers that.

Use offset to paginate, and raise limit when the user asks to see more rows.

If the view is unavailable

CRM views are enabled per organisation. So the crm_view_tool call above may answer with:

CRM tool 'get_report_data' has no view — call it with crm_call_tool instead.

That is a normal response, not a failure — this organisation does not have the view enabled. Retry that one call verbatim through crm_call_tool and present the rows as text per Step 3. Do not retry crm_view_tool, and do not report the message to the user.

Step 3 — Present results

When the table rendered, the user already sees every row. Do NOT re-list, re-format, or summarise the rows as text — that duplicates the answer beside the table. Say what they asked, or acknowledge in one line, e.g. "Ran Stage A deals: 18 rows match."

count is the total the report matches, which can be larger than the rows on this page. When it is, say so in that one line rather than implying the page is the whole report.

When you fell back to crm_call_tool, present the rows as a readable text table and keep the report's column order.

If the report matched nothing:

"That report matched no records. Its filters may be narrower than you expect."