Skip to content
Need help with our api? Ask in Community
ShipHero AI Toolkit

ShipHero AI Toolkit

ShipHero supports two ways to use AI with your ShipHero data: the ShipHero MCP and the Public API Skill.

Use the MCP if you want the simplest setup for asking ShipHero questions from an AI client. Use the Skill if you are an advanced user who wants to answer more complex ShipHero questions or if you are interested in building a ShipHero integration.

Note

Both the MCP and the Skill are read-only. They can query ShipHero data, but they cannot create, update, or delete records in your ShipHero account.

Why this exists

Warehouse operations generate constant questions: What’s the current inventory level on SKU ABC-123? Which orders have an address hold? Which shipments were created yesterday? Traditionally, answering those questions meant logging into ShipHero, running reports, or building custom integrations.

The ShipHero AI Toolkit gives AI clients a secure way to query live ShipHero data. It comprises two tools. The ShipHero MCP is the fastest path for common operational questions, while the ShipHero Public API Skill supports advanced analysis and ShipHero integration development.

OptionBest forWhere it works
ShipHero MCPSimple ShipHero lookups and operational questionsAny AI client or agent that supports MCP, including Claude, ChatGPT, and many others.
ShipHero Public API SkillComplex ShipHero questions, advanced API work, custom GraphQL queries, and integration developmentAgent coding harnesses such as Codex, Claude Code, Cursor, and many others

ShipHero MCP

The ShipHero MCP is the easiest way to connect an AI client to ShipHero. It lets supported clients query your ShipHero account without requiring you to manually write GraphQL, manage Public API tokens, or install the full skill package.

ShipHero MCP works anywhere MCP is supported, including:

ProductSupported clients
Claude ChatWeb and Desktop
Claude CoWorkWeb and Desktop
Claude CodeWeb, Desktop, and CLI
ChatGPT WorkWeb and Desktop
CodexWeb, Desktop, and CLI
Other MCP clientsAny product with MCP support

Use the MCP for questions like:

  • What’s the current inventory level for SKU ABC-123?
  • Show me orders with an address hold.
  • Show shipments created yesterday.
  • List purchase orders expected to arrive this week.
  • Show returns created today.

The MCP is designed for simple ShipHero queries. It is not intended for building full integrations or generating complex custom application code.

MCP Installation

Our MCP server URL is:

https://mcp.shiphero.com/mcp

Installation steps vary by AI client and subscription type. Follow your AI client’s documentation, or ask your AI client to help you configure the connection.

Claude

Claude calls MCP servers Connectors. See Claude’s official custom connector documentation.

ChatGPT

Use ChatGPT Plugins to connect ChatGPT on the web or in the desktop app to the ShipHero MCP. Availability depends on your plan and workspace controls.

Enable Developer mode before adding the ShipHero MCP.

See OpenAI’s official ChatGPT developer mode documentation and MCP apps in ChatGPT documentation.

Note

For individual accounts, you can add the ShipHero MCP from the Codex CLI:

codex mcp add shiphero -c mcp_oauth_callback_port=8765 \
  --url https://mcp.shiphero.com/mcp \
  --oauth-client-id 00ks8TNFn5J9SrqePsRSHKmiakONdWjU

Others

See your AI client’s documentation for MCP setup instructions. You can also ask the client to add the ShipHero MCP and point it to this page.

The ShipHero MCP server URL is https://mcp.shiphero.com/mcp.

The server requires OAuth and, where available, supports Client ID Metadata Documents (CIMD). Otherwise, use the following settings:

  • OAuth Client ID: 00ks8TNFn5J9SrqePsRSHKmiakONdWjU
  • OAuth Callback port: 41592

Alternatively, you can fall back to a stdio MCP configuration.

MCP stdio command:

npx -y mcp-remote@latest https://mcp.shiphero.com/mcp 41592 --transport http-only --static-oauth-client-info '{"client_id":"00ks8TNFn5J9SrqePsRSHKmiakONdWjU"}'

MCP Usage

After installation, ask your AI client a ShipHero question:

Get my last 50 ShipHero orders.

On first use, your browser will open a ShipHero authorization flow.

Keep prompts specific. Ask for the exact object, date range, SKU, order number, warehouse, or status you need. Specific prompts use fewer API credits and return better answers.

Public API Skill

The Public API Skill is for advanced users working from an agentic coding harness. It provides an agent with more advanced capabilities than those of the MCP option.

Use the Skill when you want an agent to:

  • Analyze ShipHero data with custom query logic.
  • Build or modify software that integrates with ShipHero.
  • Know how to use the ShipHero Public API.

The Skill requires a coding agent. It is not meant for Claude Chat or Claude CoWork type of tools. Use the MCP on those clients.

Skill Installation

Download the current Skill package:

shiphero-public-api-skill-20260604.zip

The fastest installation path is to ask your coding agent to install it:

Download the Public API Skill from https://developer.shiphero.com/shiphero-ai-toolkit and install it

For manual installation, extract the zip into your agent’s skill folder.

unzip shiphero-public-api-skill-20260604.zip -d ~/.agents/skills/

Common skill folders:

AgentSkill folder
Codex~/.agents/skills/
Claude Code~/.claude/skills/
Cursor.cursor/skills/

Restart your agent if it does not load newly installed skills automatically.

Skill Usage

After installation, you can ask your agent ShipHero questions:

How many orders did I receive today on ShipHero?

For integration work, be explicit about the target language, framework, and workflow:

Use the ShipHero Public API Skill to add a Node.js function that fetches inventory by SKU from the ShipHero GraphQL API.

The Skill includes scripts for OAuth PKCE token setup, token refresh, and token revocation. When your work is done, ask the agent to revoke the token if it no longer needs access.

API Credits

Both the MCP and the Skill query the ShipHero Public API. API calls consume API credits.

AI agents can make several calls while answering one question. Ask for only the fields and date ranges you need, and monitor credit usage during testing.