# AI agents and MCP tools for merchandise automation | MerchandAise

- slug: developers-ai-agents-mcp
- locale: en-US
- localeResolved: en-US
- canonical: https://www.merchandaise.com/en-us/developers/ai-agents-mcp

Explore MerchandAise AI agent and MCP tool discovery for custom merchandise workflows, assistant authorization, approved operations, and LLM-ready integration feeds.

## What this page covers
This is the dedicated discovery page for AI agent builders evaluating MerchandAise MCP-compatible tool access. It explains the public server card, the assistant-authorized tools list, and the launch review path.
For the human and LLM-facing product-creation promise, pair this MCP guide with /ai-custom-product-designer: describe or upload an idea, let AI design or adapt it, verify manufacturable options, and move into supplier-backed quote or configurator handoff.
Use it when an assistant needs structured access to merchandise discovery, quoting, or approved operations without bypassing human governance.

## AI agent and MCP surfaces
- **MCP server card and authorized tools:** Plan AI-agent integrations on the browser-safe MCP guide page, use the AI custom product designer pillar for intent-to-product context, then use the public server-card JSON for machine discovery and assistant-authorized endpoints for tool listing and execution. Human guide: Read MCP implementation guide (/en-us/developers/ai-agents-mcp#implementation-checklist). Machine-readable contract: Public MCP server-card JSON (/en-us/.well-known/mcp/server-card). Access: public. Public discovery document for machine clients. The tools list and tool execution endpoints still require assistant authorization at /api/v1/mcp/tools. Canonical page: /developers/ai-agents-mcp. Canonical section: /developers/ai-agents-mcp#mcp-tools.
- **Assistant tooling API docs:** Read the assistant/tooling OpenAPI docs that support design sessions, tool use, and integration reviews. Community, supplier, and enterprise contracts are listed separately. Human guide: Open assistant tooling docs (/en-us/developers/api-integrations). Machine-readable contract: Assistant tooling OpenAPI JSON (/api/docs/openapi.json). Access: public. Public machine-readable OpenAPI for assistant/tooling APIs; use the domain-specific contracts below for community, supplier, and enterprise implementation. Canonical page: /developers/ai-agents-mcp. Canonical section: /developers/ai-agents-mcp#api-docs.

## Production implementation checklist
- **Step 1: Start with the server card and authorized tools list:** Use the public MCP server card to identify the server, then request assistant authorization before reading the current tools list, expected inputs, and execution boundaries.
- **Step 2: Design assistant authorization deliberately:** Map which assistant actions can be read-only, which need user confirmation, and which require workspace or enterprise approval before execution.
- **Step 3: Test agent behavior before launch:** Validate prompts, tool calls, failure states, observability, and approval handoffs in staging before enabling live merchandise workflows.

## Production safeguards
- **Server card is public, tools are authorized:** The MCP server card is public for planning, while tool listing and execution require matching assistant authorization and workspace controls.
- **Human review for sensitive operations:** Quoting, ordering, supplier, and enterprise actions should keep approval checkpoints instead of relying only on autonomous tool calls.
- **LLM-friendly source of truth:** The page has canonical metadata, structured data, a dedicated LLM feed, and stable links so assistants can cite the current integration surface.

## Before an AI agent workflow goes live
Keep MCP server discovery separate from tools-list and execution readiness. Production agent workflows need authorization, review boundaries, and explicit human-control points.
- **Map tool access by risk:** Separate read-only discovery, quoted recommendations, supplier-sensitive operations, and order-affecting actions before enabling execution.
- **Design human confirmation paths:** Define which user, supplier, or enterprise approvals are required before an assistant can mutate merchandise, pricing, or fulfillment state.
- **Run prompt and tool-call QA:** Test refusal behavior, bad inputs, stale tool definitions, telemetry, and recovery paths before live assistant traffic reaches the MCP surface.

## FAQ
- **What is the MerchandAise MCP tools endpoint for?:** It lets authorized assistants retrieve MCP-compatible tool definitions for approved MerchandAise capabilities. The server card is public, while the tools list and execution require authorization.
- **Can an AI agent place orders automatically?:** Production agent workflows should preserve the required approval, authentication, and workspace controls before any sensitive operation is executed.
- **Is there a dedicated LLM feed for this page?:** Yes. The page exposes a focused LLM feed at /llm/developers/ai-agents-mcp with the canonical summary, tool links, safeguards, and implementation checklist.

## Planning an agent workflow?
Share the assistant use case, target users, approval model, and required tools. We will recommend the right MCP, API, and governance pattern.
- **Talk to integrations team:** /en-us/contact?intent=agent-mcp-review
- **View AI product designer:** /en-us/ai-custom-product-designer