Xianmu IMS provides a single AI-agent interface for inventory setup, stock operations, audits, alerts, purchasing, projects, and warehouse data through MCP or REST APIs.
The fastest way to install a skill directly from the registry.
npx clawhub@latest install xianmu-ims
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
Copy this prompt to OpenClaw to install it automatically.
Help me install xianmu-ims using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Get the raw skill files in a ZIP archive.
Xianmu IMS is an Openclaw Skills integration for operating a user-deployed Xianmu IMS instance on a local machine, NAS, or private network. It combines inventory search, inbound and outbound stock, transfers, adjustments, stocktakes, low-stock alerts, item records, locations, suppliers, attachments, documents, purchasing, projects, and BOM workflows behind one skill entry point.
The skill also includes a guided initialization experience for creating categories and warehouse locations. It validates connectivity, selects an onboarding scenario, previews the proposed configuration, requests confirmation, and performs administrator-only writes through MCP. This makes Openclaw Skills useful for both first-time setup and safe day-to-day inventory management without sending instance data to external services.
BASE URL and, when available, an API Token beginning with xmt_. Tokens determine viewer, member, or admin permissions and are protected by a 60-request-per-minute limit.BASE/api/mcp when the client supports remote MCP, or use the REST API at BASE/api/v1 when MCP is not practical.data field and explain API errors using the returned code, message, HTTP status, and details.Use the unauthenticated health endpoint before attempting token-based operations:
curl -s http://127.0.0.1:8066/api/v1/health
A response containing "status": "up" confirms that the instance is online. Record the reported version and edition. If the default address is unavailable, ask for the user’s instance address, for example http://<instance-host>:8066, and test it again.
If no instance responds, do not repeatedly try tokens. Follow the installation and connection guide in references/setup.md, including the supported Windows installation flow or the user’s selected deployment method. The instance may run locally, on a NAS, or on a remote private network.
In Xianmu IMS, open Settings → API Token, create a token, and provide it to the session only when needed. The token is displayed in plaintext once and typically starts with xmt_. Use the least privilege required:
viewer: read-only operations.member: normal inventory writes.admin: initialization writes, scrap outbound operations, and administrative workflows.A token is strongly recommended even in single-user mode because it enables rate-limit protection and operation auditing.
For remote MCP-capable clients, configure the Streamable HTTP endpoint:
BASE/api/mcp
Send the token as:
Authorization: Bearer xmt_<token>
The MCP endpoint exposes 67 tools, including 21 read tools and 46 write tools, subject to edition, item-count, role, and read-only restrictions. If MCP is unavailable, use REST with the same authorization header and the API prefix:
BASE/api/v1
The OpenAPI definition is available at BASE/api/openapi.json, with interactive documentation at BASE/api/docs.
For requests such as “initialize,” “create categories,” “create locations,” or “start using the system,” follow references/onboarding.md. The workflow detects the current state, selects a scenario, asks a limited set of questions, previews the proposed categories and locations, obtains confirmation, and writes the configuration through MCP using an admin token.
| Element | Structure or rule |
|---|---|
| REST base path | BASE/api/v1 |
| MCP endpoint | BASE/api/mcp using Streamable HTTP |
| Authentication | Authorization: Bearer <Token> |
| Success envelope | { "code": "OK", "message": "ok", "data": <business data> } |
| Error envelope | { "code": "<error>", "message": "<description>", "details": ... } |
| Pagination | page and page_size; server maximum 200, with 50 or fewer recommended |
| Read-only mode | Append ?mode=readonly to the MCP URL |
| Rate limit | 60 requests per minute |
items[]. Removing an item only unlinks its documents; deleting a document is a separate irreversible action.references/setup.md: connectivity checks, installation, token creation, verification, and activation guidance.references/api.md: REST endpoints, request bodies, parameters, and response examples.references/onboarding.md: the initialization lifecycle for categories and locations.references/scenarios.md: seven onboarding scenarios and default templates.references/mcp-tools.md: initialization tool names, return structures, and naming rules.AUTH_REQUIRED indicates a missing or invalid token; FORBIDDEN indicates insufficient role permissions; EDITION_REQUIRED indicates a REST or feature entitlement restriction; LIMIT_EXCEEDED indicates the free-edition 50-item limit; OUT_OF_STOCK indicates insufficient quantity; DUPLICATE_ITEM indicates a possible duplicate; NOT_FOUND indicates a missing entity; RATE_LIMITED indicates more than 60 requests per minute; and READONLY_MODE indicates that writes are disabled.
?mode=readonly for demonstrations and safer inspection environments.references/ files while keeping the single Openclaw Skills entry point maintainable.Loading
FreeCAD CADPilot connects OpenClaw to a local FreeCAD instance through CADPilot’s stdio MCP server and guides safe, advertised-tool-only CAD workflows.

A Node.js-powered Openclaw Skills integration that retrieves and organizes public Xiaohongshu notes, creator posts, comments, and engagement metrics for research and decision-making.

Infinite Creation orchestrates a local AI workflow that transforms novels, scripts, or ideas into voiced comic dramas and short videos.

Help independent creators get their manga, webcomic, webtoon, light novel or anime discovered by listing or claiming it in AnimeHunt’s free directory.

call4me lets your AI agent call businesses to handle bookings, changes, cancellations, questions, and hold times, then report back with the outcome.

A five-engine reasoning suite that routes high-stakes questions through auditable workflows, verifies results, labels confidence, and refuses unsupported conclusions.








































