MCP playground

Try MCP on a live CRM. Right now.

This is Pact's real MCP substrate — the one Claude, Cursor, or any Model Context Protocol client connects to — pointed at a sandbox workspace. Ask a question, watch the model choose tools, call them live, and cite the records it used. No signup, no email, no card.

Ask the sandbox workspace anything a rev-ops lead would. The model picks Pact MCP tools, calls them live, and cites the records it used.

Sandbox workspace · illustrative composite data · sessions are rate-limited and logged.

Everything you just did was governed.

The playground runs the production engine with a least-privilege tool roster — the guardrails aren't a demo mode, they're the product.

Sandbox tenant

You're querying an isolated workspace seeded with illustrative composite data. No real customer records anywhere in the loop.

Consent-gated

The same consent-at-selection filter production runs — tools the caller's scopes don't cover are never even offered to the model.

Every call audited

Each prompt and tool call writes the same audit rows an authenticated session would. That's the product, demonstrated.

Costs you nothing

Model spend lands on our meter, capped daily. If the playground hits its shared budget, it says so and pauses until tomorrow.

Five tools in the sandbox. Nine in the product.

The playground offers the read-only roster below. Connect your own workspace and the server adds contacts (BYOK-decrypted), agent firing, and daily briefings — see the full roster.

query_accounts
Search CRM accounts. Consent-suppressed domains are filtered out — and counted.
query_deals
Search pipeline deals, consent-filtered by the linked account.
query_pipeline_health
Aggregate pipeline metrics — open value, weighted value, win rate, by stage.
get_metric_explanation
Ask why a KPI moved. Runs the metric-explainer on aggregates only — no PII.
ask_workspace
A natural-language question across the workspace. Grounded, cited answer.

Now run it from your own AI client.

The same server you just talked to is live at https://api.pact.place/mcp/. Sign up, mint a token, and point Claude Desktop, Cursor, or the official MCP SDK at your own workspace.

Claude (native connector)

Settings → Connectors → Add custom connector, paste the endpoint, sign in to Pact and approve the read scopes — OAuth discovery handles the rest.

Full setup guide — Claude, Cursor, Python

~/.cursor/mcp.json

{
  "mcpServers": {
    "pact": {
      "url": "https://api.pact.place/mcp/",
      "headers": { "Authorization": "Bearer pact_live_xxx" }
    }
  }
}

Want this against your real CRM?

The sandbox is the appetizer. A demo gets you the full nine-tool server on your own data — consent rules, BYOK, audit trail and all.

Get a demo Start free