Deploy your agent anywhere.
Ground it on Vasco.
Claude or your own stack. Wherever your agent runs, it needs the same revenue-data substrate underneath: consistent metrics, account timeline from lead to cash, your actual plan, integrity. One MCP endpoint. 58 tools. Three layers.
Quick start · ~30 seconds
One command. You're in. Then ask anything.
1 - Install
$ claude mcp add-json vasco '{"type":"http","url":"https://mcp.vasco.app/mcp","oauth":{...}}'
> Added http MCP server vasco to local config
▐▛███▜▌ ▝▜█████▛▘ ▘▘ ▝▝
Claude Code v2.1.145
Opus 6.7 (1M context) · Claude Team
~/revenue-ops
❯ /mcp
⎿ Authentication successful. Connected to vasco.
⎿ ✓ 58 tools loaded · acme-corp · scopes: read:metric, read:context, read:plan
2 - Ask
❯ What's the win rate for expansion this quarter, and which at-risk accounts should we prioritize?
⏎ Thinking…
⎿ query-metric-engine expansion_win_rate, period=2026Q2 → 0.42 (n=12)
⎿ query-metric-engine expansion_win_rate, period=2026Q1 → 0.38 (n=13)
⎿ get-plan-forecast expansion arr_target=$90k → $82.5k booked (+2pt vs pace)
⎿ query-context-graph at_risk_expansion · signal=champion_change in_last=60d → 3 accts
Expansion · Q2 2026 · win rate 42% (+4pt QoQ)
On track on dollars: $82.5k of the $90k target booked and win rate up to 42% (+4pt QoQ) — fewer but larger deals (5 of 8 by count).
Red flag: Q3 SAL pipeline is critically light at 33% coverage; 3 accounts also show champion-change in the last 60 days.
───
Accounts to prioritize
1. Heliox — HIGH
Champion left in March; renewal in 45d with no exec sponsor identified. Multithread to a new owner this week.
2. Nimbus — MEDIUM
Usage flat two quarters; expansion talk stalled on a pending SDK auth fix. Unblock to reopen the conversation.
↑ Stratos — EXPANSION-READY
Usage up 3× this quarter, added 2 seats, asked about the enterprise tier. Strong upsell to close this month.
───
Bottom line: Q2 dollars are on track — 42% win rate (+4pt QoQ), $82.5k of $90k booked on fewer, bigger deals. The red flag is Q3 pipeline at 33% coverage — get ahead of it now. Multithread Heliox before its renewal and close the Stratos upsell.
REQUIRES: Authenticated Vasco user · Trial or paid workspace · Sources connected: onboarded
Onboarding takes ~10 min · we wire reconciliation for you.
Ready to get started?
Vasco works with any MCP-compatible client. Pick yours below and you're connected in under a minute.
# Get your clientId and callbackPort from app.vasco.app after sign-up.
# OAuth opens in browser. Reads auto-approved · writes need confirmation.
$ claude mcp add-json vasco '{"type":"http","url":"https://mcp.vasco.app/mcp","oauth":{"clientId":"$VASCO_CLIENT_ID","callbackPort":$VASCO_CALLBACK_PORT}}' # ~/Library/Application Support/Claude/claude_desktop_config.json
# Restart Claude Desktop. OAuth runs on first tool call.
{
"mcpServers": {
"vasco": {
"transport": "http",
"url": "https://mcp.vasco.app/mcp",
"oauth": "true"
}
}
} # ~/.cursor/mcp.json (works at workspace level too)
{
"mcpServers": {
"vasco": {
"command": "npx",
"args": ["-y", "@vasco/mcp", "--remote"],
"env": { "VASCO_OAUTH": "interactive" }
}
}
} import { MCPClient } from "@modelcontextprotocol/sdk";
const mcp = new MCPClient({
url: "https://mcp.vasco.app/mcp",
auth: { kind: "oauth2.1", scope: "read:metric read:context" }
});
await mcp.connect();
const tools = await mcp.listTools(); // 58 tools across 3 layers
await mcp.callTool("get-metric-engine-schema", {});
const r = await mcp.callTool("query-metric-engine", {
metric: "win_rate",
motion: "enterprise",
period: "2026Q2"
});
// → { value: 0.283, definition_id: "win_rate_v3", n_deals: 47, citations: [...] } curl -X POST https://mcp.vasco.app/mcp \
-H "Authorization: Bearer $VASCO_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"method": "tools/call",
"params": {
"name": "query-metric-engine",
"arguments": {
"metric": "win_rate",
"motion": "enterprise",
"period": "2026Q2"
}
}
}' What you'd have to build
Six categories of work between your agent and a real answer.
Each is its own engineering project. Tap through to see what each layer actually contains — and what it costs to build before you have one.
Total realistic build: ~9 months for v1.
Vasco ships this on day one. One MCP endpoint. Same data layer underneath every agent we've built and every one you'll build.
End-to-end pipeline
Sources in. Reconciled. Layered. Served.
Sources
Vasco · Reconcile
Layers
MCP Server
Clients
Three layers · one coherent answer
The revenue data architecture, exposed as MCP.
Every tool call falls into one of three layers — and they compose. The example above used all three: Foundation for the actuals, Plan for the gap, Context for the at-risk accounts. Pick any layer and the others fall in behind.
"What is the business? How does it perform?"
GTM Model
Stages, channels, motions, functions, employees, dimensions, journeys.
Source Mappings
CRM → GTM entity mapping. Preview-before-save. Versioned.
Integrity Radar
Data quality score, issues, rules, auto-reconciliation.
Metric Engine
Actuals, trends, conversions, breakdowns. Deterministic.
"Where is the org headed? How is it tracking?"
Goals & Scenarios
Targets, what-if plans, scenario diffs, apply & rollback.
Plan Inputs
Conversion rates, quotas, costs, retention, channel spend.
Forecasts
Projected revenue, pipeline, customers — by month, by motion.
Benchmarks
Industry comps at P25 / P50 / P75.
"Why do the numbers look the way they do?"
Context Graph
Entities, relationships, account context. Traverse to identify at-risk, segment, or explain churn.
Graph Schema
Entity types, relationship types, queryable attributes. The intended first call before any traversal.
Saved Queries
Named slices of the graph — reusable across sessions, citable as [^context-N].
Artifacts
MetricQuery, ContextQuery, Reports, Slides — persisted, citable evidence inside agent answers.
Why this works for GTM teams
The same data layer, governed by RevOps.
Builders connect once. Then every rep, every agent, every seat reasons against the same definitions, the same plan, the same graph. RevOps governs what the agent sees — without re-platforming the org.
More than 50 tools across 3 layers. Full reference in our docs.
Every tool — name, signature, scope, response shape — lives in one place. Browse the foundation, plan, and context tools, plus how artifacts and citations work.