Talk to your data. Stay in Claude.
Ask Claude about your Brand Score, marketplace shelf, orders, Coach recommendations — in real-time, without leaving Claude.ai, ChatGPT, or Claude Code. Zobrx is the first marketing platform with a live MCP server. Query your actual data (not training cutoff), get conversational answers, stay in flow.
Watch Zobrx MCP in action — querying marketplace data from Claude
Six conversations you can start right now
Real examples of queries the Zobrx MCP answers. Claude fetches your live data, structures it, and explains what it means.
Settlement & revenue
"What was my net settlement on Snapdeal last week?"
Claude fetches your ecommerce orders, channel fees, TCS/TDS, and net settlement — breaks it down per channel and per week.
Shelf availability
"Which of my top 10 SKUs are OOS on Blinkit Mumbai right now?"
Claude queries Q-Radar pincode-level OSA — shows you which SKUs are out of stock where, instantly.
Brand Score trends
"Compare my Brand Score on Amazon vs Flipkart this month."
Claude pulls your Brand Score history per channel, shows the deltas across all 22 dimensions, and surfaces regressing areas.
Coach recommendations
"Summarize my open Coach recommendations and what to act on first."
Claude lists your cross-platform rules + anomalies, prioritizes by revenue impact, and links the action surfaces.
SKU mapping status
"Which of my SKUs are not yet mapped on Meesho?"
Claude searches your master catalog, filters by mapping status, and shows the gap.
Regional price analysis
"Show me my price corridor on Blinkit across metro vs tier-2 pincodes."
Claude queries regional shelf data and shows price distribution by region and time window.
First-to-market advantage in marketing platforms
Linear, Stripe, Notion, Intercom, PayPal, Canva, HubSpot all ship official MCP servers. But Zobrx is the first marketing or marketplace intelligence platform with a live MCP server.
The Zobrx MCP isn't a chatbot trained on docs or a one-way API. It's a live bridge: Claude asks your Zobrx workspace questions in real-time, gets current data, and reasons about it conversationally. You stay in Claude.ai or ChatGPT. No tabs, no logins, no context-switching. Your actual data, your actual AI assistant, one conversation.
This is table stakes arriving. Being here first closes a moat.
Six technical foundations that set us apart
Live data, not training cutoff
Claude queries your Zobrx workspace in real-time. Not pre-trained knowledge or last month's snapshot — your actual data right now.
Stay in Claude.ai / ChatGPT
No context-switching to web dashboards. Custom connector in Claude.ai settings, or `claude mcp add` in Claude Code. One login, all your data.
Multi-cluster tenant isolation
Zobrx data lives across 3 MongoDB planes (tenant, eVanik orders, Q-Radar shelf). Single tenant-scope wrapper secures all 3 with no extra plumbing.
RBAC + audit
Your workspace roles map to tool scopes. Every query audited and logged for SOC 2 / DPDP compliance — stronger audit trail than the web UI.
Read-only + safe by design
v1 is read-only across 10 tools. No secrets in responses. Response-capped at 10K tokens. Write tools planned for M4 behind explicit gating.
Works everywhere
Claude.ai custom connectors, ChatGPT developer mode, Claude Code, Claude API via MCP SDK, VS Code Copilot, Cursor, Gemini CLI — any spec-compliant client.
Multi-cluster, single wrapper, zero new compute
Your Zobrx data lives across three MongoDB planes: tenant (your workspace), eVanik (cross-marketplace orders), and Q-Radar (quick-commerce shelf intel). A single tenant-scope wrapper in the MCP router secures all three clusters with the same RBAC as the web UI. No sticky sessions, no new infrastructure. Just a stateless HTTP endpoint that rides your existing API server.
- Stateless Streamable-HTTP transport — forward-compatible with MCP v2
- Tenant-scoped queries across 3 Mongo planes with 1 context wrapper
- RBAC reuse — your workspace roles are tool scopes
- Audit-logged per query — stronger trail than the web UI
- Response-capped at 10K tokens — Claude-safe
- ~$0 to operate — no new servers, no new costs
User asks a question
Stateless HTTP · Verify JWT · runWithTenantContext() · Dispatch tool
Dashboard, Brand Score, catalog
Cross-marketplace orders
Pincode OSA, Leak Ledger
Structured, reasoned, actionable
From beta to full feature (M1–M4)
Where we are and where we're going.
M1 — Plumbing pilot
✅ Live (2026-06-11)Stateless MCP router, static-bearer auth, 6 tools (dashboard, integrations, catalog, Brand Score, orders, shelf OSA), feature-flagged behind MCP_ENABLED. End-to-end verified from Claude Code.
M2 — OAuth
🔄 In progressWorkOS AuthKit MCP toggle, RFC 9728 protected-resource discovery, JWKS verification, scope→permission mapping. Test from Claude.ai custom connector.
M3 — v1 launch
🔄 In progressFull 10-tool set, per-tenant rate limiting, audit-event emission, response-size guards, public docs at /docs/mcp, beta with 2–3 friendly tenants.
M4 — Expand
📋 PlannedChatGPT compatibility (search+fetch polish), Anthropic connector-directory submission, write tools (gated), internal/ops MCP for staff observability queries.
Common questions
Be among the first to ask Claude about your marketplace data.
Request beta access now. M3 (v1 public launch) is rolling out in June 2026. Until then, we're onboarding hand-picked customers to validate the tool set and get feedback.
- ✓ 14-day free trial
- ✓ No credit card
- ✓ Cancel anytime