Overview
The Demostack MCP Server lets your team ask natural-language questions about Demostack activity directly from Claude, ChatGPT, Gemini, and other MCP-compatible AI tools - without leaving the AI workspace they already use.
Sales leaders, AEs, SE leaders, admins, CS, PMM, and partner marketing can get instant answers on presenter activity, prospect engagement, demo performance, and content usage - grounded in real Demostack data, scoped to your organization.
Key facts:
Included for every Demostack customer at no additional cost.
Scoped to your organization's authenticated identity. No cross-org or benchmark data.
Setup
Connecting Demostack to your AI tool is a two-step process: your LLM admin enables the connector (https://mcp.demostack.com/mcp) for the workspace, then each user connects their own account using their Demostack login.
Step 1 - LLM Admin: Enable the Demostack connector
Note: This step is performed by the admin of your AI tool (e.g. Claude workspace owner, ChatGPT enterprise admin), not by a Demostack admin.
In your AI tool, navigate to the admin or workspace settings page where third-party connectors are managed.
Locate the Connectors (or equivalent) section and grant permission for users in your workspace to add the Demostack connector.
Add the Demostack MCP server (https://mcp.demostack.com/mcp) to the list of approved connectors. Once approved, the Demostack connector becomes available for every user in your workspace to enable individually.
The exact steps and naming vary by AI tool. Refer to your AI tool's connector documentation for tool-specific guidance.
Step 2 - Each user: Connect with your Demostack login
Once your LLM admin has enabled the connector, each user connects their own Demostack account:
In your AI tool, go to Settings → Connectors (or the equivalent integrations menu).
Find Demostack in the list and click Connect.
You will be redirected to a Demostack login page. Sign in using your Demostack account credentials.
Approve the connection. You will be redirected back to your AI tool - fully connected.
That's it. You can now start asking questions about your Demostack data directly from your AI tool.
Multi-org users: If you belong to more than one Demostack organization, the MCP will resolve your org context automatically based on your authenticated identity. You don't need to specify an organization with each query.
Alternative: Connect with an API key
If you're using a CLI or a config-file-based client, you can authenticate with a Demostack API key instead.
Generate a key: In Demostack, go to Settings → API Tokens, create a token, and copy it.
Then add the Demostack server to your tool's config:
Claude Code (CLI) - edit ~/.claude.json:
{
"mcpServers": {
"demostack": {
"type": "http",
"url": "https://mcp.demostack.com/mcp",
"headers": { "Authorization": "Bearer ${DEMOSTACK_API_KEY}" }
}
}
}
Claude Desktop - edit claude_desktop_config.json:
{
"mcpServers": {
"demostack": {
"command": "npx",
"args": [
"-y", "mcp-remote", "https://mcp.demostack.com/mcp",
"--header", "Authorization: Bearer ${DEMOSTACK_API_KEY}"
]
}
}
}
OpenAI Codex (CLI & Desktop) - edit ~/.codex/config.toml:
[mcp_servers.demostack] url = "https://mcp.demostack.com/mcp" bearer_token_env_var = "DEMOSTACK_API_KEY"
What you can ask
The MCP launches with a focused question set covering presenter activity, prospect engagement, demo performance, and content usage. Below are examples of what you can ask today.
Examples:
Which demos have the most views?
Which shared links have more than three unique visitors?
Which demos haven't been used in the last 90 days?
If a question isn't yet supported, the MCP will tell you directly instead of guessing or failing silently. Coverage will expand based on our roadmap and customer feedback.
Outcomes by team
Sales leaders walk into 1:1s and pipeline reviews with answers, not spreadsheets.
AEs prep for calls by asking who from the account has engaged, and how.
Admins keep the demo library clean by surfacing unused or underperforming content in seconds.
CS teams answer customer questions on the fly without context-switching.
L&D and PMM see which people are engaged with training assets and specific features.
Partner marketing identifies which partners use which demos most.
FAQs
Who is this for?
Any Demostack user who wants fast analytics insights without pulling a report or logging into Demostack. It's especially useful for admins, sales engineering leaders, and sales leaders.
How much does it cost?
Nothing. It's included for every Demostack customer.
Which AI tools does it work with?
Optimized for Claude (Desktop and Code). Compatible with ChatGPT, Cursor, and any other MCP-compatible client.
OAuth or API key - which should I use?
OAuth is simplest for in-app clients (Claude Desktop, ChatGPT) and ties to your Demostack login. API keys suit CLI tools and automated or config-driven setups.
Does the MCP support the latest Claude, ChatGPT, and Gemini models?
Yes. It works with all current Claude, Gemini, and ChatGPT models, including Claude Opus 4.7 and GPT-5.5.
Does the model I use affect the answers I get?
Answer quality may vary by model. Frontier models like Claude Opus 4.7 and GPT-5.5 produce the strongest results, while lighter models trade some depth for speed. The MCP confines possible outcomes rigidly regardless of the model.
Can it change anything in my Demostack account?
No. The MCP is read-only.
Is my data shared across organizations?
No. All access is scoped to your organization's authenticated identity. There is no cross-org or benchmark data.
Can a bad prompt damage my Demostack data or workspaces?
No. The MCP is fully read-only, so it cannot create, edit, or delete anything.
What if my question isn't supported yet?
The MCP launches with a focused question set. If a question isn't supported, the MCP will tell you directly instead of guessing. Coverage will expand based on our roadmap and customer feedback.
What's the difference between the MCP Server and the Analytics API?
The MCP is for questions you want to ask, in plain English, inside your AI workspace. The Analytics API is for systems you want to build: custom agents, internal dashboards, automations, and pipelines. Both are read-only, scoped to your org, and free.
I belong to multiple Demostack organizations. How does that work?
The MCP resolves your org context automatically from your authenticated identity. You don't need to specify an organization with each query.
Troubleshooting
The Demostack connector doesn't appear in my AI tool.
Your LLM admin has not yet enabled it for your workspace. Reach out to your AI tool's workspace admin and ask them to complete Step 1 above.
The MCP returns no data for a question I expected to work.
Check that your Demostack account has the permissions required to view the data you're asking about. The MCP respects your existing Demostack permissions.
My answers seem outdated or inconsistent.
Try rephrasing the question more specifically (e.g. include a time window or asset name). If the issue persists, share the question with your CSM so we can improve coverage.
The LLM fails to answer.
As this is still in beta, we may release updates from time to time. Try disconnecting and re-connecting Demostack's connector on your LLM.