DemandIntel in Claude, Codex & your tools

Model Context Protocol

Your platform, inside the tools you already think in.

Connect DemandIntel to Claude, Codex, Cursor or any MCP client with a single command, the installer drops in our training Skill and wires up the MCP server in one hit. Ask it what to coach, and let it build the training.

No account yet? Request beta access, once approved you can generate your token and connect in under a minute.

Claude CodeClaude DesktopCodexCursorContinueChatGPT desktop…any MCP client

The loop

Query. Plan. Create. Deliver. Review - then round again.

DemandIntel is not a one-off report generator. Wired into your AI tool it becomes a loop where every engagement makes the next one sharper - because the review that closes one cycle is the evidence that opens the next.

1
Query
Ask DemandIntel, in plain language, what's really happening on the calls - no dashboards to hunt through.
2
Plan
It answers from real QA and competency data, and points to the one highest-leverage gap to fix.
3
Create
Your AI tool turns that gap into ready-to-run training: session plan, notes, handouts, scripts, scenarios.
4
Deliver
You run the session - grounded in the team's own conversations, not a generic curriculum.
5
Review
Weeks later, ask again: did the score move? The answer is the evidence - and the next question.

A worked example

From one question to a delivered training session - live.

Here is the loop running end to end on a real demo account, exactly as it plays out in Claude. No exports, no slide deck to assemble by hand - the tool does the listening and the prep; you stay the expert in the room.

“Use DemandIntel to analyse Acme Co’s training requirements for the last week. What is the most important training need where I can impact the team’s performance? Then create the sales-training assets I need to deliver the training.”
You, in Claude - to the DemandIntel MCP

Step 1

Query

You, in Claude or Codex:

“Analyse Acme's last week through DemandIntel. What's the one training need I can fix to move the team's performance?”

Natural language → the ask() tool + QA data.

Step 2

Plan

DemandIntel answers with evidence:

Reps book meetings without confirming budget.

QA data confirms it's the whole team, not one rep. Fix first: qualification.

Step 3

Create materials

Claude builds the training kit:

60-minute session plan
Trainer's notes
Handouts & scripts
Practice scenarios

Built from the team's real calls, via a Skill.

Step 4

Deliver

You run the session:

Deliver the training to the team, grounded in their own conversations, not a generic template.

The trainer stays the expert; the tool preps.

Step 5

Review

Two weeks on, ask again:

“Did qualification scores move?”

Pull the QA trend and see the change, evidence of impact.

The result becomes the next question. ↺

Every answer sharpens the next question — a virtuous circle

The same query, re-run two weeks later, becomes the proof the training worked — and the start of the next cycle.

Create

Your AI tool turns that gap into ready-to-run training.

Session plan, trainer notes, handouts, practice scripts and scenarios, all built from your team's real calls. You stay the expert in the room; the tool does the prep.

Session plan with timings and objectives

Trainer notes and talking points

Handouts and reference cards

Practice scripts from real calls

Scenarios based on actual objections

Assessment questions to check retention

Get the full training kit with DemandIntel Skills

Skills are downloadable documents that turn your platform data into complete course materials. One Skill contains every asset you need for a single training topic, ready to deliver. Sign up to access the Skills library.

Get connected

One command. Skills and MCP, wired up.

No config files to hand-edit, no OAuth dance. Mint a token, paste one line into your terminal, and the installer drops the DemandIntel training Skill into your assistant and registers the MCP server (~31 tools) in a single hit.

Step 1

Mint a token

In Settings → API & MCP Access, click Create token. Choose read scope, optionally cap monthly usage or lock it to specific clients. Copy the di_pat_ secret, shown once.

Step 2

Run the one-liner

Paste the installer command below into your terminal with your token. It installs the DemandIntel Skill into ~/.claude/skills/ and registers the MCP server automatically.

Step 3

Ask

Restart your assistant. DemandIntel and its Skill are now available, ask a question in plain language and your tool picks the right DemandIntel tool automatically.

Run this in your terminal

One command, Skill installed, MCP server registered.

curl -fsSL https://raw.githubusercontent.com/demandintel/demandintel-skills-resource/main/install.sh | DEMANDINTEL_API_KEY=di_pat_… bash

Replace di_pat_… with the token from Step 1. The installer needs curl and bash, standard on macOS and Linux. Then restart your assistant and run /mcp in Claude Code to confirm demandintel is live.

Using Cursor, Codex, or another IDE? Show the config snippets.

~/.cursor/mcp.jsonCursor / IDEs

{
  "mcpServers": {
    "demandintel": {
      "url": "https://mcp.app.demandintel.io/v1?token=di_pat_…"
    }
  }
}

~/.codex/config.tomlCodex (stdio via mcp-remote)

[mcp_servers.demandintel]
command = "npx"
args = ["-y", "mcp-remote", "https://mcp.app.demandintel.io/v1", "--header", "Authorization: Bearer di_pat_…"]

Prefer to install the Skill manually? Download the pack and drop it into your assistant's skills folder.

What you can reach

Around 31 tools - plus one that answers anything.

Curated tools return instantly. ask handles the genuinely novel questions in natural language, routing to the right data behind the scenes.

Free-form Q&A

ask

Spot-check & coaching

top_objectionsmost_improved_repcoaching_plansrep_league_tabledisposition_trendaccuracy_trend

Competency & QA

competency_heatmapqa_scoresget_competency_scoresdrift_alertslist_frameworks

Reports, calls & entities

list_reportsget_reportget_report_recommendationslist_callscall_statslist_clientslist_reps

All tools are read-scope today; mutations - for example generating reports - require a write-scoped token.

Safe by design

Scoped, logged, and revocable in a click.

Token, not password

Access is a scoped Personal Access Token; only a hash is stored, the secret is shown once and can be revoked instantly.

Scoped to what's needed

Read-only by default; an optional client filter limits a token to specific clients - ideal for a contractor or partner.

Every call accounted for

Each request is logged with status, duration and usage, so you can see what a token has done and cap its spend.

Set an optional monthly usage cap on any token as a spend control - the platform stops the token once the cap is reached.

Close the loop

Ask your platform what to coach - and let it build the session.

The listening, the analysis, and the prep, inside the tool you already use. You bring the judgement and the room.

Need an account? Request beta access to connect the DemandIntel MCP server.