How Do You Connect Flux to Claude or Cursor?

Edge

Flux exposes 54+ MCP tools that let you query your LinkedIn data, score drafts, map engagement graphs, and run prospecting workflows directly from Claude Desktop, Cursor, or any MCP-compatible client. You get conversational access to the same analytics engine that powers the Flux dashboard — ask a question in plain language, and the AI calls the right tools behind the scenes. Setup takes about two minutes: grab your API key from Settings and add the endpoint to your client's MCP config.

What you'll learn

Setting up the Flux MCP connection in Claude Desktop or Cursor
The eight tool clusters and what each one does
Four ready-to-use workflow recipes: content analysis, target account engagement, meeting prep, and draft scoring
Which tools cost credits and which are free graph reads
Chaining tools together for multi-step analysis

How do you set up the connection?

You need two things: an Edge-tier Flux account and an MCP-compatible client (Claude Desktop, Cursor, or similar).

Get your API key. Open Flux Settings and copy your API key. This key authenticates every MCP call and ties it to your account's credit balance.
Add the MCP endpoint to your client. In your client's MCP configuration file, add an entry pointing to Flux's MCP server. The endpoint is your Flux app URL at /api/mcp, and the API key goes in the authorization header.

For Claude Desktop, add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "flux": {
      "url": "https://fluxgraph.app/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

For Cursor, the same structure goes in your MCP settings. Once configured, your client discovers all available Flux tools automatically — you can start asking questions immediately.

What tools are available?

Flux's 54+ tools are organized into eight clusters. You do not need to memorize them — your AI client discovers them automatically and picks the right one based on your question. But understanding the clusters helps you ask better questions.

LinkedIn — Live data pulls from public profiles. Ask Claude to pull someone's recent posts, look up a company's activity, or search for people by keyword. These use credits because they fetch fresh data from LinkedIn.
Posts — Query your stored post library. Search by meaning, find similar posts, get posting insights. Useful for content audits and repurposing.
Score — Draft scoring calibrated against your own history, connection scoring for outreach warmth, and angle suggestions for new content.
Traversal — Read-only queries against the engagement graph: who engaged, from which companies, how often. These are zero-credit database reads.
Streams — Create and manage recurring saved searches that surface new matching posts on a schedule.
Segments — Build and query saved audience slices — groups of people you want to track or target.
Collections — Manage saved post lists for reference, repurposing, or competitive analysis.
Account — Check your credit balance, usage history, and billing status.

How do you analyze your last 30 days of content?

This is the most common starting point. Ask Claude something like: "Analyze my LinkedIn posting performance over the last 30 days."

Behind the scenes, Claude calls get_posting_insights which returns your posting frequency, engagement trends, top-performing posts, and content type breakdown — the same data that Flux benchmarks across 25,150 posts from 66 tracked profiles. You get a clear picture of what formats and topics drove the most engagement, with multipliers compared against your own median.

Follow up with specifics: "Which of my posts got the most comments?" or "Show me my best-performing carousels." Claude chains the right tools together without you needing to name them.

How do you find who is engaging from target companies?

This is where Flux earns its keep for GTM teams. Ask: "Who from Acme Corp has been engaging with my posts?"

Claude uses get_post_engagers_by_company to group your engagers by employer, surfacing the specific people at target accounts who reacted or commented. If three people from a prospect company engaged with your last post about a problem you solve, that is a buying signal worth acting on today.

To go deeper, ask: "Who consistently engages with my content across multiple posts?" Claude runs get_post_set_engagers across your recent posts and identifies repeat engagers — the people who keep coming back. Cross-reference those against your target account list and you have a warm pipeline ranked by engagement frequency. These traversal queries are zero-credit reads, so you can run them as often as you want.

How do you prep for a meeting in 30 seconds?

You have a call in 20 minutes. Ask Claude: "Brief me on Sarah Chen before my meeting."

Claude chains two tools: get_profile_summary pulls Sarah's recent posting activity, topics, format preferences, and engagement patterns. Then get_engagement_brief shows whether Sarah or anyone at her company has engaged with your content — which posts, how recently, and how often. You walk into the call knowing what she cares about and whether her team is already warm.

For account-level prep, ask: "Show me all engagement from people at Acme Corp in the last 60 days." The engagement brief scoped to a company gives you every touchpoint across your content — specific people, specific posts, specific dates.

How do you score and improve a draft before posting?

Paste a draft into your chat and ask: "Score this post and tell me how to improve it."

Claude calls score_draft, which evaluates your draft against your own last 200 posts — not generic best practices, but what actually works for your specific audience. The score returns a predicted engagement multiplier relative to your median performance, plus dimensional feedback on hook strength, format fit, topic relevance, and length.

The real workflow is iterative: score, revise based on the feedback, score again. Each round refines the draft against your calibrated baseline. Ask "What would make this a 2x post?" and Claude applies the scoring feedback to suggest specific edits.

Which tools are free to use?

Traversal tools — the engagement graph queries — are zero-credit database reads. This includes get_post_engagers_by_company, get_post_set_engagers, get_engagement_brief, get_company_engagement_overlap, get_alumni_engagers, and the other graph-traversal tools. These read data Flux has already collected, so they cost nothing to run.

Tools that fetch fresh data from LinkedIn (the linkedin_* cluster) use credits because they pull live information. Your Edge plan includes 200 credits per month, and you can check your balance anytime by asking Claude "How many credits do I have left?"

FAQ

Do I need to know the tool names to use Flux MCP? No. Describe what you want in plain language — "show me who engaged from target accounts" or "score this draft" — and your AI client picks the right tool. The tool names in this guide are for reference, not for memorization.

Does MCP work with clients other than Claude Desktop and Cursor? Yes. Any client that supports the MCP protocol can connect to Flux's endpoint. The setup is the same: point the client at https://fluxgraph.app/api/mcp with your API key in the authorization header.

Can I chain multiple tools in a single conversation? Yes, and this is where MCP shines. Ask "Brief me on this prospect and then score a draft reply" and Claude calls get_profile_summary, get_engagement_brief, and score_draft in sequence, using each result to inform the next step.

Is my data shared with Claude or other AI providers? Flux's MCP server returns data to your local AI client. The data flows through the same channel as any other MCP tool call — your client's privacy settings and data handling apply. Flux does not send your data to third-party AI providers independently.

What plan do I need for MCP access? MCP tools require an Edge-tier subscription ($15/month), which includes 200 credits per month, AI scoring, semantic search, multi-profile support, and full API access.

On this page