propose_unfollow — LinkedIn MCP tool reference

What it does

Undo a follow.

This tool drives a real LinkedIn account

It runs through a connected sender account in your workspace — not through a scraper and not through your own profile. Call list_sender_accounts first to see which identities are connected and healthy. If the workspace has none, this tool returns an error telling you so; accounts for agents covers how teams get one without risking their own profile.

Arguments

Name	Type	Required	Meaning
profile	string	yes	Profile URL, public identifier (the /in/ slug) or provider id
reason	string	no	Why, in one or two sentences — shown to the person who approves it
account	string	no	Sender account id or name to act as. Defaults to the healthiest connected account.
dry_run	boolean	no	Check everything (target, do-not-contact, duplicates, text, limits) and return the verdict without proposing anything

This table is generated from the literal inputSchema that tools/list returns, so what your client sees and what this page says are the same object:

inputSchema

{
  "type": "object",
  "properties": {
    "profile": {
      "type": "string",
      "description": "Profile URL, public identifier (the /in/ slug) or provider id"
    },
    "reason": {
      "type": "string",
      "description": "Why, in one or two sentences — shown to the person who approves it"
    },
    "account": {
      "type": "string",
      "description": "Sender account id or name to act as. Defaults to the healthiest connected account."
    },
    "dry_run": {
      "type": "boolean",
      "description": "Check everything (target, do-not-contact, duplicates, text, limits) and return the verdict without proposing anything"
    }
  },
  "required": [
    "profile"
  ]
}

Request

Paste this, swap the key, and you have called the tool.

curl

curl -s https://topclozer.com/api/mcp \
  -H "Authorization: Bearer tcz_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "propose_unfollow",
      "arguments": { "profile": "ACoAAB1x…" }
    }
  }'

Returns

{ approval_id, status, duplicate, kind, target, warnings, account, next }. With dry_run: { dry_run: true, would_propose: true|false, refusal?, target, checks: { do_not_contact, duplicate, text, limit } } and nothing is queued.

The payload arrives inside an MCP content block — one text block whose body is this JSON, pretty-printed. Parse the string; there is no structured-content field.

payload

{
  "approval_id": "apr_lx3k2a9fq",
  "status": "pending",
  "duplicate": false,
  "kind": "comment",
  "target": { "label": "Post by Dana Whitfield", "id": "urn:li:activity:7332661864792854528" },
  "warnings": [],
  "account": "Marcus Feld",
  "next": "Waiting for a person in the TopClozer console → Approvals. Poll get_approval."
}

Worth knowing

Nothing reaches LinkedIn from this call. A person approves (and may edit) it on the console's Approvals page; it then goes out under the account's Smart Limits, one action per account per pass, inside working hours. The do-not-contact list is checked here and again at send time.

Rate limit

30 LinkedIn tool calls per 5 minutes, per workspace. Exceeding it returns HTTP 429 with JSON-RPC error -32002 and a Retry-After header. The budget is shared by every LinkedIn tool and counted per workspace, so minting a second key does not buy a second allowance.

The limit exists because an agent retrying in a loop is the request pattern that gets a LinkedIn account restricted, and a restricted account is a much worse outcome for you than a 429. Workspace tools do not spend this budget.

Related tools

propose_follow

Propose following a LinkedIn member — a softer first touch than an invitation. NOT done until approved.

list_following

Members and pages the connected account follows

All 65 tools are listed on the tool index. Connecting a client is on the Claude setup page, and the protocol details are on the server page.