Recipes / Weekly cold email review: rank your Instantly campaigns and fix the weak ones

Weekly cold email review: rank your Instantly campaigns and fix the weak ones

This recipe uses the official Instantly MCP server, and it works in Claude. Steps that write, send or spend credits wait for your approval.

IntermediateSDRs Marketing

Who is this for?

SDRs, SDR managers and marketers who run cold email in Instantly. You get a short weekly read on what's working, what's broken and what to change, without clicking through every campaign. The same idea works with Smartlead, lemlist or Reply.io.

What do you get?

Last 7 days, ranked, with one fix waiting for approval. Sample output: every company and person in it is fictional.

Instantly campaigns, last 7 days. The analytics don't say whether replies include out-of-office auto-replies, so read reply rate next to positive replies.

Campaign	Sent	Reply rate	Positive replies	Bounce rate	Call
Ops leaders, freight (Larkspur-style accounts)	1,240	4.1%	9	0.8%	Working. Keep it
RevOps, mid-market SaaS	980	2.6%	4	1.1%	Fine. Step 2 gets most replies
Founders, Series A	1,510	0.7%	0	6.4%	Needs attention
Re-engage closed-lost	310	0.0%	0	0.3%	Not sending since Tuesday

What stands out: 1. 'Founders, Series A' has a 6.4% bounce rate. The list likely needs verifying before more sends. 2. 'Re-engage closed-lost' stopped sending Tuesday. Its one sending account ([email protected]) is paused. 3. In 'Ops leaders, freight', step 1 ('Quick question about dock scheduling') drives 7 of the 9 positive replies. Worth testing that opener in the RevOps campaign. Waiting for approval: pause 'Founders, Series A'. Nothing has been changed yet. All campaigns, accounts and numbers here are fictional. The sender is Quarry Lane Analytics (fictional).

How long does it take, and what does it cost?

Estimate About 10–15 minutes a week, most of it spent deciding what to change.

Assumptions: One weekly run over 5–10 campaigns and their sending accounts.

Instantly: Reads each campaign's analytics for the last 7 days and checks sending-account status. After you approve, it makes one pause or update per campaign you pick. Credit use not confirmed by Instantly.

Exact credit costs depend on your plan with each vendor. Check the tool pages before a large run.

Which MCP servers does it use?

Every server is confirmed in: ClaudeCursor

Instantly Up · auth

ClaudeCursorRead + writeCredits not confirmed
https://mcp.instantly.ai/mcp

How do I set it up in a chat app?

Every server in this recipe is confirmed for Claude. Connect each one, then paste the prompts below into one conversation.

Connect Instantly without code (Claude)

Developer setup

Auth setup and mcp.json for Claude Code and Cursor

Instantly: Paste an API key from the vendor's settings into the config (header or URL). There's no browser login, so it works well headless.

Put this in .mcp.json (Claude Code, project root) or ~/.cursor/mcp.json (Cursor). Swap the example keys for your own, and don't commit real ones. Remove the // comment lines if your client requires strict JSON.

{
  "mcpServers": {
    "instantly": {
      "url": "https://mcp.instantly.ai/mcp",
      "headers": { "Authorization": "YOUR_INSTANTLY_API_KEY" }
    }
  }
}

Step-by-step prompts

Pull analytics for all my active Instantly campaigns for the last 7 days: sent, replies, reply rate, positive replies and bounce rate. Tell me whether replies include auto-replies. Show it as a table ranked by positive replies, and mark any campaign with fewer than {min_sends} sends as 'too early'. Don't change anything.
For each campaign, check whether it sent every day this week and whether its sending accounts are active or paused. List any campaign that stopped sending or has a bounce rate over {bounce_limit}%.
For the top campaign, show which sequence step gets the most replies (if the analytics break replies down by step), and quote that step's subject line. Based on it, suggest one test I could run in a weaker campaign. Just describe the test. Don't set anything up.
Give me up to three recommended actions, such as pausing a campaign or fixing a paused sending account. Wait for my approval on each one. Then make each approved change in Instantly, one at a time, and show me the result.

Can I run it on a schedule? (optional)

Run it weekly at the same time (Monday morning works), so each week compares like for like.
In a scheduled run, have it read the analytics and account status, write the report, then stop. You approve every pause or change yourself.
Run it by hand the first time. That shows you what a normal week looks like before you trust its 'needs attention' calls.

Where does this go wrong?

Out-of-office replies inflate the reply rate. Auto-replies can count as replies. Ask whether the numbers include them. Compare reply rate with positive replies before you call a campaign a winner.
Judging a campaign on too few sends. A campaign with 80 sends and one reply isn't a trend. Set a minimum (for example 300 sends this week). List smaller campaigns as 'too early' instead of ranking them.
A sending problem mistaken for bad copy. A sudden drop to zero replies often means the campaign stopped sending. Check each campaign's sent count and the status of its sending accounts before anyone suggests a copy change.
Pausing the wrong campaign. Campaign names can look alike ('Founders A' vs 'Founders A v2'). Before pausing, show each campaign's exact name, status and last send date. Pause one at a time.

Gotchas

Instantly's server uses an API key, not a sign-in. In Claude, you add it as a custom connector with the key inside the URL. Treat that URL like a password and don't share it.
Instantly's docs say the API, and so the connector (its MCP server), needs the Growth plan or higher.
The server can also reply to emails, change workspace settings and manage API keys and billing. Keep those out of this recipe, and use a key with the narrowest access you can.
Its listing confirms it works in Claude and Cursor. ChatGPT support isn't confirmed yet.

Download

Get the prompts, guardrails and expected output as one Markdown file. Paste it into a Claude Project or ChatGPT project's instructions, or save it as CLAUDE.md for Claude Code.

Download instantly-weekly-campaign-review.md

Vendor sites: Instantly. If we join an affiliate program, these links will be labeled as affiliate links.