Skip to main content

How to Use Modjo MCP

Practical guide to install, configure, and use the Modjo Model Context Protocol (MCP) with your AI assistant. Discover available tools, real-world use cases, and best practices.

Summary

The Modjo Model Context Protocol (MCP) allows your AI assistant (Claude Desktop, Cursor, etc.) to directly access your Modjo conversational and CRM data to automate call, deal, and account analysis. This guide explains how to install, configure, and use Modjo MCP daily.

Target audience: Modjo platform administrators

Prerequisites

Before starting, make sure you have:

  1. An active Modjo account with administrator access

  2. A Modjo API key (see Configuration section below)

  3. An MCP-compatible client: Claude Desktop, Cursor, Dust, n8n, Zapier, Make, Codex CLI, Gemini CLI

Required permissions: Administrator role in Modjo

Plan: Feature available on all plans (check with your CSM)

⚠️ Important note: Modjo MCP is currently in public beta. Endpoints, tool names, and behaviors may evolve without notice.

API Key Configuration

Step 1: Retrieve your Modjo API key

  1. Log in to your Modjo workspace

  2. Navigate to Settings → Integrations

  3. Locate the Modjo API Key section

  4. Click Generate a key (or copy the existing key)

  5. Immediately copy and store this key in a secure password manager

⚠️ Security: This key grants access to all your workspace data. Never share it or commit it to a Git repository.

Step 2: Configure Claude Desktop

To use Modjo's MCP with Claude Desktop:

  • You must have an Admin access and access to the Developer settings

  • You must have a working installation of Node.js on your computer. If that's not the case, you can follow this Node.js installation guide

Configuration:

  1. Open Claude Desktop

  2. Navigate to Settings → Developer → Edit Config

  3. Add the following block to the JSON file:

{
  "mcpServers": {
    "modjo": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.mcp.modjo.ai/v1/mcp",
        "--header",
        "Authorization:${MODJO_AUTH_HEADER}"
      ],
      "env": {
        "MODJO_AUTH_HEADER": "Bearer your-modjo-api-key-here"
      }
    }
  }
}

  1. Replace your-modjo-api-key-here with your API key copied in step 1

  2. Save the file

  3. Completely restart Claude Desktop (quit the application and relaunch)

⚠️ Windows Tip: Claude Desktop has a bug on Windows where spaces in arguments are misinterpreted. Use no space after the colon in the --header argument (Authorization:${VAR}) and place the full value Bearer <key> in the env block.

Available Tools

Once MCP is configured, your AI assistant has access to 12 Modjo tools grouped into 4 categories:

1. Search tools (Discovery & scoping)

  • modjo_mcp__get_accounts: List/search accounts (CRM ID, CRM link, etc.)

  • modjo_mcp__get_deals: List/search deals (amount, dates, status, etc.)

  • modjo_mcp__get_calls: List/search calls (filters by deal, account, contact, date, user…)

  • modjo_mcp__get_contacts: List/search contacts (CRM person ID, email, phone, job title…)

  • modjo_mcp__get_users: List/search internal users (ID, role, department…)

  • modjo_mcp__get_emails: List/retrieve related emails (content included)

2. Content tools

  • modjo_mcp__get_transcript: Retrieve the transcript of one or more calls (timestamped blocks + speaker)

3. AI analysis tools (agent-first)

  • modjo_mcp__ask_anything_on_call: Ask an AI question about a call (with Modjo agent option)

  • modjo_mcp__ask_anything_on_deal: Ask an AI question about a deal (with Modjo agent option)

  • modjo_mcp__ask_anything_on_account: Ask an AI question about an account (with Modjo agent option)

4. Agent catalog

  • modjo_mcp__get_agents: Discover available agents (Modjo or custom)

Practical Use Cases

Use Case 1: Prepare a meeting in one line

Situation: You have a call with Contoso account in 15 minutes.

Command in Claude:

"Brief me on Contoso, last calls, open deal, sentiment, next steps"

Result: Your assistant uses ask_anything_on_account with the MeetingPrepper agent and returns a structured brief in seconds.

Use Case 2: Analyze weekly pipeline

Situation: Monday morning, you want to identify at-risk deals.

Command in Claude:

"Which of my Q2 open deals haven't moved in 10 days and had a pricing objection in the last call?"

Result: MCP lists the relevant deals, then executes ask_anything_on_deal on each to extract objections.

Use Case 3: Instant post-call coaching

Situation: You just finished a demo and want immediate feedback.

Command in Claude:

"What did I do well, what did I miss, and what should I follow up on?"

Result: One ask_anything_on_call with the SalesCoach agent provides structured feedback in 20 seconds.

Use Case 4: Draft a personalized follow-up email

Situation: You need to send a follow-up email after a demo with Jean.

Command in Claude:

"Draft a follow-up email to Jean after the demo, referencing the integration concern he raised at minute 14."

Result: The EmailFollowUp agent reads the transcript and drafts a personalized email.

Use Case 5: Detect churn signals (CSM)

Situation: You manage 20 accounts and fear weak churn signals.

Command in Claude:

"Look at my top 20 accounts. Find the ones where the last three calls sounded lukewarm and tell me why."

Result: One ask_anything_on_account per account with the ChurnInspector agent surfaces at-risk accounts.

Use Case 6: Quarterly win/loss analysis

Situation: You lost 7 deals in Q1 and want to understand why.

Command in Claude:

"I lost 7 deals last quarter. What reasons came up in the close calls?"

Result: MCP retrieves lost deals, calls ask_anything_on_deal with the DealBriefing agent, and clusters reasons.

Best Practices

  1. Always prefer ask_anything_on_* over raw transcripts
    ❌ Bad: Retrieve transcript and ask Claude to analyze it
    ✅ Good: Use ask_anything_on_call with the appropriate agent

  2. Choose the right agent for the right task
    Use get_agents to discover available agents.
    Native Modjo agents: CallQualifier, CallSummary, DealBriefing, EmailFollowUp, MeetingPrepper, NextStepper, SalesCoach, ChurnInspector

  3. Limit the scope of heavy questions
    ❌ Bad: "Analyze all calls from this deal over 6 months"
    ✅ Good: "Analyze the last 5 calls from this deal"

  4. Don't use MCP for bulk statistics
    MCP is designed for qualitative analysis, not mass counting.

  5. Secure your API key
    Never share it or commit it to Git. If leaked, immediately revoke in Settings → Integrations.

  6. Use environment variables
    Never hardcode the key in versioned config files.

Limitations and Constraints

  1. Read-only access: MCP cannot modify data in Modjo or your CRM

  2. Timeouts on large deals/accounts: ask_anything_on_deal and ask_anything_on_account may timeout after 60 seconds

  3. No native aggregated statistics: You must retrieve calls via get_calls and calculate client-side

  4. Near real-time data: Transcripts appear within 5 minutes after a call ends

  5. Tenant-level authentication: Anyone using your API key can access all calls, deals, and accounts

Troubleshooting

Issue: Server not appearing after config change

Fix:

  1. Completely restart your MCP client (quit the application, not just close the window)

  2. Verify the JSON file is valid (no missing comma, closed braces)

Issue: Authentication errors (401/403)

Fix:

  1. Verify your API key is correct in the env block

  2. Verify the format is Bearer your-api-key

  3. Test the variable: echo $MODJO_AUTH_HEADER (Unix) or echo %MODJO_AUTH_HEADER% (Windows)

  4. Regenerate a new key if needed

Issue: Connection timeouts

Fix:

  1. LLM clients often enforce a 60-second limit for external tools

  2. Limit your question scope ("based on the last 5 calls")

FAQ

Q: I'm not technical. Can I really set up MCP?

A: Yes. If you can paste a password into a settings screen, you can connect MCP to Claude Desktop in 5 minutes.

Q: Will MCP update my Salesforce?

A: No, not directly. MCP reads Modjo only. To write elsewhere, pair it with n8n, Zapier, or your data team's workflow.

Q: Can I get "calls per rep this quarter" via MCP?

A: Technically yes, practically no. For anything that'll end up on a chart or slide, use the Modjo API.

Q: Can I change what Modjo agents return?

A: Native agents have fixed prompts. But you can create custom agents in the Modjo interface and call them via MCP by their UUID.

Q: Is the data real-time?

A: Close to it. Transcripts appear within 5 minutes after a call ends.

Q: Someone leaked my API key. Now what?

A: Immediately revoke it in Settings → Integrations. The new key works instantly.

Q: I'm already using API v2. Should I migrate to MCP?

A: No. They complement each other. Keep the existing pipeline and add MCP for conversational use cases.

See Also

  • Modjo API v2 Documentation (for bulk data pipelines)

  • Complete guide "Modjo MCP and API for AEs and CSMs" (internal link to add)

  • Tutorial "Create a custom Modjo agent" (internal link to add)

Last updated: 2026-05-06

Related Articles
Manually import a call/meeting into Modjo
Our AI Subcontractor: Transparency and Compliance
Discover AI Insights in Modjo
How Modjo x HubSpot works
What's New in Modjo
Did this answer your question?