Skip to main content

Modjo MCP — Getting Started Guide

Connect your AI to your Modjo data

The Modjo MCP connects your AI assistant directly to your Modjo data: calls, deals, accounts, and contacts. Ask a question in plain language from your usual assistant and get an answer in seconds, without switching tools.

⚠️ Public beta. Endpoints, tool names, and behaviors may change without notice. We recommend against relying on the MCP for business-critical production workflows at this stage.

Prerequisites

Before getting started, make sure you have the following:

  • An active Modjo account with Administrator role

  • A Modjo API key — available from your Settings → Integrations

  • A compatible MCP assistant, meaning one of the following applications that will serve as your interface to interact with your data: Claude Desktop, Cursor, Dust, n8n, Zapier, Make, Codex CLI, or Gemini CLI

The feature is available on all Modjo plans. If in doubt, reach out to your Customer Success Manager.

Step 1 — Get your API key

  1. Log in to your Modjo workspace

  2. Go to Settings → Integrations

  3. Find the Modjo API Key section

  4. Click Generate a key (or copy the existing key if one has already been created)

  5. Copy and store your key immediately in a secure password manager

⚠️ Modjo does not store your API key and will not be able to retrieve it for you later. Copy it as soon as it is generated. This key provides access to all data in your workspace — never share it or commit it to a code repository.

Step 2 — Configure your AI assistant

The Modjo MCP server is available at https://api.mcp.modjo.ai/v1/mcp using the Streamable HTTP protocol, authenticated via Bearer token.

Configuration varies depending on the application you use. Follow the relevant section below.

Claude Desktop

Claude Desktop only supports stdio transport. To connect to a remote HTTP server like Modjo, a mcp-remote proxy is required to bridge the two.

Prerequisites: Administrator access to Claude Desktop and Node.js installed on your machine.

  1. Open Claude Desktop and go to Settings → Developer → Edit Config

  2. Add the following block to your claude_desktop_config.json file:

json

{   "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

  2. Save the file

  3. Fully quit Claude Desktop (Cmd+Q on macOS) and relaunch it

Verification: A hammer icon should appear at the bottom right of the chat input area. Click it to confirm that the Modjo tools are listed.

Windows users: Due to a Claude Desktop bug on Windows, do not include a space after the colon in the --header parameter. Place the full Bearer <key> value in the env block where spaces are handled correctly.

Dust

Dust natively supports remote MCP servers through its interface — no command line required.

  1. In your Dust workspace, go to Spaces → Tools → Add Tools

  2. Click Add MCP Server

  3. Enter the server URL: https://api.mcp.modjo.ai/v1/mcp

  4. For authentication, select Bearer Token and paste your Modjo API key

  5. Dust will automatically sync the available tools

  6. Assign Modjo tools to your agents from the Agent Builder

n8n

Since version 1.88.0, n8n natively supports MCP — no community package required. Two modes are available.

Mode A — with an AI Agent (recommended)

  1. Add an AI Agent node with your LLM credentials (OpenAI, Anthropic, etc.)

  2. Add an MCP Client Tool node and connect it to the Tools input of the Agent

  3. Configure the node:

  4. Under Tools to Include, select All or choose specific Modjo tools

Mode B — standalone (deterministic step)

  1. Add an MCP Client node to your workflow

  2. Configure:

  3. Select the tool to run from the auto-populated dropdown

  4. Map input parameters and connect to downstream nodes

Other compatible clients (Claude Code, Cursor, Zapier, Make…)

For any MCP client that supports header-based authentication, use the following generic configuration:

json

{   "mcpServers": {     "modjo": {       "type": "http",       "url": "https://api.mcp.modjo.ai/v1/mcp",       "headers": {         "Authorization": "Bearer your-modjo-api-key-here"       }     }   } }

The 12 available tools

Once configured, your assistant has access to 12 tools across four categories.

Search and discovery

Tool

Description

get_accounts

Search for accounts (the name filter is required)

get_deals

Search for deals (status, amount, dates, loss reason…)

get_calls

Search for calls (by deal, account, contact, date, user…)

get_contacts

Search for contacts (email, phone, job title…)

get_users

Search for internal users (role, department…)

get_emails

Retrieve associated emails, including content

Transcripts

Tool

Description

get_transcript

Full transcript of one or more calls, with timestamps and speaker identification

AI-powered analysis

Tool

Description

ask_anything_on_call

AI question on a call, with optional Modjo agent

ask_anything_on_deal

AI question on a deal, with optional Modjo agent

ask_anything_on_account

AI question on an account, with optional Modjo agent

Agent catalog

Tool

Description

get_agents

List available agents (native Modjo or custom)

Modjo agents

The MCP is built around an agent-first approach: rather than asking open-ended questions, you select a specialized agent that produces structured, consistent output.

Native agents

Agent

UUID

Best for

CallSummary

741e9ffc-87be-4bca-bb0d-f167be8b963e

CRM-ready call summary

NextStepper

09715241-0cdd-44c9-a386-92a1340bdf4a

Action items and commitments (your tasks vs. theirs)

MeetingPrepper

c0a76fd7-3f56-4a03-b2e7-7765da10c457

Pre-meeting brief + questions to ask

CallQualifier

3a7753e1-d21e-4cb7-a990-7820291274cd

Structured qualification + gaps

DealBriefing

1204e84f-6edd-4782-bbdf-e5e070b400cf

Executive deal summary + risks

EmailFollowUp

b2a9ae1b-2026-4dfa-9c67-101733a87a04

Personalized follow-up email draft

Custom agents are also available: CallScorer, ChurnInspector, SalesCoach, MEDDPICC Validator, QuickBANT, and more. Use get_agents to discover them.

Usage examples

Prepare a client meeting "Brief me on Contoso — last calls, open deal, sentiment, next steps" The MeetingPrepper agent generates a structured situation summary before your call.

Identify at-risk deals "Which Q2 deals haven't moved in 10 days and had a pricing objection on the last call?" The MCP identifies the deals and automatically extracts the objections raised.

Get post-demo feedback "What did I handle well, what could I improve, and what needs follow-up?" The SalesCoach agent produces structured feedback in under 20 seconds.

Write a personalized follow-up email "Draft a follow-up email for Jean referencing the integration issue raised at minute 14." The EmailFollowUp agent reads the transcript and drafts a context-aware message.

Analyze reasons for lost deals "I lost 7 deals last quarter. What reasons came up in the closing calls?" The MCP retrieves the relevant deals and groups the causes by theme.

Best practices

Use agents instead of raw transcripts. Using ask_anything_on_call with the right agent is consistently more effective than analyzing a raw transcript. Reserve get_transcript only when verbatim quotes are strictly necessary.

Check the existing summary first. get_calls returns a pre-generated summary field for most calls. Read it before calling ask_anything_on_call for basic questions — it may already have the answer.

Scope your requests precisely. Specify "over the last 30 days" or "since the last call" to avoid analyzing an unnecessarily broad history.

Use get_agents proactively. Before asking a generic question, check whether a specialized agent exists for your use case (coaching, competitive analysis, churn detection…).

Respect pagination. Tools return a maximum of 50 results per page. Use limit: 50 and iterate with nextCursor when exhaustive results are needed.

Set the language when relevant. For French-speaking users or French data, add language: "fr" to your ask_anything_on_* calls to get responses in French.

Limitations

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

  • Strict date format: dateRange filters require the format YYYY-MM-DD (e.g. 2026-03-12). Any other format will cause an error.

  • No native aggregate stats: for volume metrics, use the Modjo API.

  • Pagination required: search tools are limited to 50 results per request. Use nextCursor to iterate.

  • AI analysis latency: ask_anything_on_deal and ask_anything_on_account may time out after 60 seconds — narrow the scope if this happens.

  • Tenant-scoped access: anyone using your API key has access to all calls, deals, and accounts in the workspace.

Troubleshooting

The MCP server doesn't appear after configuration Fully quit the application (don't just close the window) and relaunch it. Also verify that your JSON file is correctly formatted — check for missing commas or unclosed brackets.

Authentication error (401 or 403) Check that your API key is correctly entered and that the format is Bearer your-api-key. Test with echo $MODJO_AUTH_HEADER (Unix) or echo %MODJO_AUTH_HEADER% (Windows). If the issue persists, generate a new key.

Request timeout Reduce the scope of your query — for example, limit analysis to the last 5 calls rather than the full history.

Security

  • Never hard-code your API key in a versioned configuration file

  • Never share your key — anyone who has it can access all data in your workspace

  • If your key is compromised, revoke it immediately from Settings → Integrations. Revocation is instant and a new key can be generated without any downtime.

FAQ

I'm not technical. Is the setup accessible? Yes. If you can copy-paste a key into a settings screen, you can complete the setup in under 5 minutes.

Can the MCP update my data in Salesforce or another CRM? No. The MCP is read-only. For write actions to other tools, combine it with n8n, Zapier, or Make.

Can I customize native agent responses? Native agents have fixed prompts. However, you can create custom agents directly in the Modjo interface and call them via their UUID in ask_anything_on_*.

How do I get aggregate statistics (e.g. number of calls per month)? The MCP has no dedicated stats tool. Retrieve calls via get_calls filtered by period and calculate metrics client-side, or use the Modjo API v2.

Is the data real-time? Near real-time. A slight delay may exist between the end of a call and its availability via the MCP.

My API key was compromised. What should I do? Revoke it immediately from Settings → Integrations. The new key is active instantly.

I'm already using the Modjo API v2. Do I need to migrate? No, the two are complementary. Keep your existing pipeline and use the MCP on top of it for conversational interactions.

See also

Related Articles
Modjo Analytics
Manually import a call/meeting into Modjo
Setup your Modjo
Introduction to Modjo Deals
Discover Modjo's Features
Did this answer your question?