PullPush MCP — connect any two APIs and keep them in sync
The official MCP server for PullPush.ai — an integration platform that connects e-commerce stores, warehouses, marketplaces, ERPs, and POS systems, and keeps stock levels, orders, and products in sync across all of them.
This server lets AI agents (Claude, Cursor, or anything that speaks Model Context Protocol) build, run, and monitor those integrations end to end: connect two systems, preview the sync safely, and cut over to live — without a human clicking through a dashboard.
Why agents like it
- Deterministic runtime, no LLM in the hot path. AI helps at build time (connector generation, field-mapping suggestions); the sync engine itself is predictable, versioned, and observable.
- Safe by default. New flows start in shadow mode (read-only).
pullpush.syncaction:"preview"shows a structured source-vs-destination diff before anything is written. - Credentials never touch your machine or env. Connection secrets are envelope-encrypted (AES-256-GCM, per-tenant keys) server-side the moment they're created.
- Self-describing. A
get-startedMCP prompt walks from zero to a working integration;pullpush://resources expose the tool catalog, quickstart, and agent guide — no pre-briefing needed. - Structured, actionable errors. Failures return
{code, suggestion, suggestedTool, retryable}— an agent always knows the next step. - Server-grade reliability. Circuit breakers, per-key ordered delivery, idempotency, dead-letter queue with retry, exponential backoff.
Try it in 60 seconds — no signup
curl -X POST https://api.pullpush.ai/api/sandbox
Returns a free 48-hour sandbox tenant: an API key, ready-to-paste MCP configs, and two
pre-wired demo connections whose stock data deliberately disagrees — so your first
pullpush.sync action:"preview" shows a real canonical diff (drifted quantities, a missing
item, in-sync rows). The sandbox deletes itself on expiry — or keep it:
POST /api/sandbox/claim (Bearer = the sandbox key, {"email": "..."}) converts it to a
persistent free tenant with everything you built preserved.
Quick start
No clone or install needed — run directly via npx:
{
"mcpServers": {
"pullpush": {
"command": "npx",
"args": ["-y", "kaduno-pullpush-mcp@latest"],
"env": {
"PULLPUSH_API_KEY": "<your-api-key>"
}
}
}
}
Add this to ~/.cursor/mcp.json (Cursor), Claude Desktop's config, or the MCP config for your
client. Prefer a remote server? Connect straight to the hosted endpoint instead:
{
"mcpServers": {
"pullpush": {
"type": "streamable-http",
"url": "https://api.pullpush.ai/mcp",
"headers": { "Authorization": "Bearer <your-api-key>" }
}
}
}
Then ask your agent to run the get-started prompt — it walks through the whole flow:
catalog connectors → create + test connections → suggest a field mapping → create the flow in
shadow mode → preview the diff → cut over.
Getting an API key
For a throwaway key, use the sandbox above. For a persistent tenant:
- Sign up free (1 connection, 1,000 events/month — no card)
- Admin panel → Settings → API Keys
- Create a key with the scope you need (tenant, org, or admin)
What's inside
45 tools (28 explicit tools plus 173 actions across 17 domain routers) covering the full integration lifecycle:
- Connections — create, test, and configure links to external systems (23+ built-in platforms: Shopify, WooCommerce, Magento, Amazon, Linnworks, Xero, HubSpot, …); any REST, GraphQL, or MCP API via the AI connector builder
- Flows — wire source → destination sync pipelines in canonical data models (stock levels, products, orders)
- Sync — preview, reconcile, and monitor; structured diffs in canonical space
- Shadow mode — run new integrations read-only until proven, then cut over
- Diagnostics —
pullpush.explaindiagnoses failures with evidence and suggested fixes - Ops — health dashboards, anomaly detection, dead-letter recovery
- Workflows — multi-step DAG automations with branches, delays, and lookups
- Import/Export — file-based data in and out
Full tool reference: pullpush.ai/docs/mcp
Environment variables
| Variable | Required | Default | Description |
|---|---|---|---|
PULLPUSH_API_KEY | Yes | — | Your PullPush API key |
PULLPUSH_API_URL | No | https://api.pullpush.ai | API base URL (override for self-hosted) |
Auth scopes
| Scope | Access |
|---|---|
| Tenant key | Your tenant's connections, flows, sync, diagnostics |
| Org key | All tenants in your organization (read or read+write) |
| Admin key | Full platform access (billing, partners, all tenants) |
Tools are automatically scoped: a tenant key never sees another tenant's data.
Features
- Auto-refresh: tool catalog refreshes every 5 minutes — new server-side tools appear without a restart
- Offline quickstart: built-in
pullpush://docs/quickstart-localresource works without network - Retry on unknown tool: automatically refreshes the catalog when a tool is not found
- Clear error messages: auth failures, network issues, and version mismatches are explained
Documentation
- Getting started — 5 minutes to a first integration
- MCP guide — full tool reference and client setup
- Shadow mode · Preview & reconcile · Security
- llms.txt — machine-readable product summary
- Machine-readable integration catalog
- MCP registry entry:
ai.pullpush/pullpush-mcp
Also available: pullpush-mcp-dev
For connector authors, platform admins, and internal developers, pullpush-mcp-dev exposes additional tools for:
- AI-powered connector generation and composition
- Canonical schema management
- Marketplace publishing
- Billing and partner administration
- Batch operations across all tenants
Development
git clone https://github.com/Kaduno-systems/pullpush-mcp.git
cd pullpush-mcp
npm install
npm run dev # Run from source via tsx
npm run build # Compile with tsup
This repository is a published mirror of the MCP client package; the platform itself is developed in a private monorepo. Issues and questions are welcome here or at hello@pullpush.ai.
License
MIT