# Search Contacts
> Searches for and retrieves contact (people) records from the Carta CRM. Use this skill when the user says things like "find a contact", "search contacts", "look up a person", "show me contact details for [name]", "full details on [name]", "tell me about [name]", "get contact by ID", "list contacts", "find people at [company]", "search people", or "/search-contacts". Returns contact details including ID, name, email, title, company, and tags. The contact ID returned can be used with the update-contact skill.
- Skill: `carta/search-contacts` (Agent Skill)
- Install (CLI): `npx skillmds@latest add carta/search-contacts`
- Raw SKILL.md: https://api.skillmd.com/api/skills/carta/search-contacts/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: carta (https://skillmd.com/u/carta)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/carta/search-contacts
---
`_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": [":", ...], "model": "", "surface": ""}`
`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.
## Overview
Look up contacts in the Carta CRM. A request about **one named person** renders that
person's card; a request for a **set** renders a table. Route on that distinction first —
it decides every call below.
## Step 1 — Determine intent: one person, or a set?
- **Detail** — the user named one person and wants the record: "full details on Jane Doe",
"tell me about Ihar", "more on this contact", "who is Jane Doe". → Step 2.
- **List** — the user wants a set, or filtered or plural results: "people at Acme",
"contacts in my pipeline", "list contacts". → Step 3.
- **By ID** — the user gave a contact ID → Step 2, skipping the resolve.
A named single person is a **detail** request even when the user says "search" or "find".
If it's genuinely unclear, treat it as a list and ask what they want to narrow to.
## Step 2 — Detail: resolve the name, then render the card
**Resolve through `crm_call_tool`, never `crm_view_tool`.** This step is for you, not the
user: a view call collapses every array in the response to a count, so the rows — and the
`id` you need — never reach you, and the user gets a list they did not ask for.
```
crm_call_tool({
"name": "crm:search_contacts",
"arguments": { query: "", limit: 10 }
})
```
Then branch on how many candidates came back:
- **Exactly one match** → render its card and stop:
```
crm_view_tool({ "name": "crm:fetch_contact_by_id", "arguments": { id: "" } })
```
- **Several matches** → do NOT guess. Render the candidates as a view and ask which one:
```
crm_view_tool({
"name": "crm:search_contacts",
"arguments": { query: "", limit: 10 }
})
```
Then ask: "Several contacts match — which one did you mean?" When they pick, call
`fetch_contact_by_id` for it. Namesakes are common in a CRM, so opening the top hit
unasked shows the wrong person with full confidence.
- **No match** → say so; do not render an empty view.
When the user gives an ID outright, there is nothing to resolve — one call, one card.
Render at most one card per request. If the user named several people, ask which to open
rather than stacking views.
## Step 3 — List: search and render the table
```
crm_view_tool({
"name": "crm:search_contacts",
"arguments": {
query: "",
limit: 20
}
})
```
If the user mentions a specific list or folder by name, resolve the name to a list ID
first, then pass `list_id` to narrow the search. This lookup has no view of its own, so
it goes through `crm_call_tool`:
```
crm_call_tool({ "name": "crm:get_contact_lists", "arguments": {} })
```
Increase `limit` if the user asks to see more results. Use `offset` to paginate.
### If the view is unavailable
CRM views are enabled per organisation, and single-record views behind a second flag on
top of that. So any `crm_view_tool` call above may answer with:
> CRM tool 'search_contacts' has no view — call it with crm_call_tool instead.
That is a normal response, not a failure — this organisation does not have that view
enabled. Retry that one call verbatim through `crm_call_tool` and present the result as
text per Step 4. Do **not** retry `crm_view_tool`, and do not report the message to the
user.
A detail request whose card has no view still resolves the same way: keep the
`crm_call_tool` resolve from Step 2 and present the chosen record as text.
## Step 4 — Present results
**When a card rendered**, the user sees the whole record. Do not restate its fields.
Answer what they asked, or acknowledge in one line.
**When a table rendered**, the user already sees every row. Do NOT re-list, re-format, or
summarise them as text — that duplicates the table. Answer the question they actually
asked, or acknowledge in one line (e.g. "Found 23 contacts — the ID is in the first
column, for `/update-contact`.").
**When you fell back to `crm_call_tool`**, display all non-empty fields in a readable
summary — name, title, company, email, phone, and tags — and show the ID prominently,
since the user needs it to run `/update-contact`.
`fetch_contact_by_id` also returns related deals and notes. The view renders those, but
call them out in text if the user is asking for context on a specific person.
If no contacts are found:
> "No contacts found matching your search. Try a different name, email, or keyword."