Icebreakly
FeaturesHow it worksPricingFAQMCPWhitelabelAffiliateLog in

Model Context Protocol

63 comment-campaign tools,
inside your AI assistant.

Connect Claude, Cursor, or any MCP client to your Icebreakly workspace. Launch always-on comment campaigns, pick who to monitor, approve AI-drafted comments, tune your Voice, and read your warming pipeline — in plain language, with no dashboard in the way.

Set up in ClaudeStart free 7-day trialSpeak to the founder
MCP serverhttps://mcp.icebreakly.com

Identical in the web app and the desktop app: no config file and nothing to install. Claude hands you off to Icebreakly’s own authorize page, you paste the key once, and the tools appear in the chat.

  1. 1

    Create an API key in Icebreakly

    Integrations → Claude Connector (MCP) → New Key. Copy the key — and note the server URL, you need it in step 3.

  2. 2

    In Claude, open Connectors → Add custom connector

    Settings → Connectors, then the Add menu in the top right.

  3. 3

    Name it Icebreakly and paste the server URL

    Name: Icebreakly. URL: https://mcp.icebreakly.com. Leave Advanced settings alone and press Add.

  4. 4

    Paste your API key and authorize

    Icebreakly’s own authorize page opens. Paste the key from step 1, press Authorize, and Claude comes back connected. Ask it to "audit my whole Icebreakly account" to confirm the tools are live.

Available tools

63 tools. Your whole warming engine.

These are Icebreakly's own tools, in the same groups the Claude Connector screen shows you when you connect. All of them are reachable in natural language — your assistant picks the right one, so there are no parameters to learn and no docs to read.

Comment campaigns

8 tools
  • create_comment_campaignSpin up an always-on comment campaign: a Voice, a schedule, and monitor mode.
  • update_comment_settingsSwitch approval vs autopilot, comment style, and whether to like each post.
  • activate_campaignValidates your LinkedIn sender and settings, then flips the campaign live.

Monitored leads

6 tools
  • add_monitor_targetsAdd the LinkedIn profiles to watch — the campaign comments on their fresh posts.
  • remove_monitor_targetsStop watching a profile without touching the rest of the list.
  • set_post_target_watcherTune how a specific post or profile is monitored and re-checked.

Comments & approvals

6 tools
  • get_pending_commentsEvery AI-drafted comment waiting on you, each with its touch number.
  • approve_pending_commentApprove a draft and queue it to post — with the post reaction if that is on.
  • update_pending_commentEdit the wording before it goes out, or deny it entirely.

Warming & pipeline

4 tools
  • get_warming_leadsEvery monitored lead by warmth stage — Stranger to Warm — ranked by touches.
  • get_campaignsYour campaigns with live comment stats: posts found, commented, liked.
  • get_campaigns_summaryOne roll-up across every campaign, so you can read the whole engine at a glance.

Voices

6 tools
  • get_promptsYour saved Voices — the tone, length, and guardrails each campaign writes in.
  • update_promptTune a Voice: tone, style notes, emoji, language, sample comments, guardrails.
  • test_comment_voiceDraft sample comments in a Voice on real posts before you go live.

Dashboard & analytics

3 tools
  • get_dashboard_summaryComments posted, posts found, posts liked — today and all-time — plus warming.
  • get_recent_postsThe fresh posts your campaigns just found from monitored leads.
  • get_performance_analyticsTrends over time so you can see the warming engine building.

AI setups

7 tools
  • get_ai_setupsThe business context and ICP every campaign and Voice is built from.
  • update_ai_setupEdit what you do and who you serve — the ground truth for your comments.
  • generate_ai_setup_icpDraft the ideal-customer profile for a setup from a short description.

LinkedIn accounts

15 tools
  • get_linkedin_accountsWhich LinkedIn accounts are connected, and their health.
  • validate_linkedin_connectionCheck an account before a campaign leans on it to comment.
  • reconnect_and_check_linkedin_2faReconnect a dropped account and walk through 2FA.

Workspace & jobs

8 tools
  • get_tasksEvery queued and running job — post fetches, comment sends — with status.
  • retry_taskRe-run a job that failed, without redoing the whole campaign.
  • get_account_statusPlan, limits, and integration health for the workspace behind the key.

Just ask

No parameters. Plain English.

Your assistant reads the intent from the prompt and calls the right tool. Every example below is a real request — the tool it lands on is on the right.

  • Start a comment campaign that watches my target accounts and drafts a comment on every fresh post.create_comment_campaignmonitor mode, always on
  • Add these 12 LinkedIn profiles to what my warming campaign watches.add_monitor_targetsnew leads to monitor
  • Show me the comments waiting on me, and how many touches each lead has.get_pending_commentseach with its touch number
  • Approve the top three drafts and like the posts too.approve_pending_commentqueues the comment + reaction
  • Who is warm and ready for me to reach out to?get_warming_leadsranked by touches, by stage
  • Draft three sample comments in my Voice so I can hear the tone.test_comment_voicebefore you go live
  • Turn on autopilot and like every post my campaign comments on.update_comment_settingsreactions on, no approval queue
  • Why is my warming campaign not posting any comments?diagnose_campaignthe recorded reason, plus the fix

More on the AI side of the platform: the Claude integration, MCP for LinkedIn sales, and every Icebreakly integration.

Common questions.

What is MCP?

MCP, the Model Context Protocol, is an open standard that lets AI assistants call external tools directly. Instead of pasting exports into a chat, your assistant calls Icebreakly in real time and gets back structured, live data from your own workspace — then acts on it: start a comment campaign, approve the drafts waiting on you, or check who is warm and ready.

Which AI clients are supported?

Any MCP-compatible client. The server speaks streamable HTTP, so Claude.ai, Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Cline, and Zed all connect. Config for each is in the section above.

Do I need an API key?

One key, created in Icebreakly under Integrations → MCP. Claude connects by sign-in: you add the server URL, Icebreakly opens its own connect page, you paste the key once and press Allow. File-config clients take the same key as an Authorization: Bearer header. Deleting the key revokes access immediately.

Is it safe to give an AI access to my LinkedIn?

Access is scoped to the one workspace behind that key, and it runs through the same Icebreakly layer the dashboard uses, so the same posting limits and safety rules apply. The assistant never touches your LinkedIn password, your client asks you to confirm before any comment or reaction posts, and the endpoint is rate limited to 300 requests per minute.

Can I use this in an automated agent workflow?

Yes, and it is a good fit. Anything that speaks MCP works: n8n, Make, custom Claude agents, or your own pipeline. The tools return structured JSON, and the long-running ones (fetching fresh posts, drafting comments) hand back a task ID you can poll.

Does MCP access cost extra?

No. It is part of the platform rather than a separate add-on, so it comes with your plan. The same per-plan limits that apply in the dashboard — monitored leads, comments per day — apply here too, at the same rate.

63 tools. One URL.

Create a key, paste the server URL, and let your assistant run the warming for you. Free for 7 days, no credit card required to start.

Start free trialSee plans and pricing