Cisco IQ MCP Server
A local MCP server that exposes Cisco IQ's Assets and Assessments REST APIs as read-only MCP tools, enabling AI assistants to query asset inventory, contracts, lifecycle data, security advisories, and field notices.
README
Cisco IQ MCP Server
A local Model Context Protocol (MCP) server that exposes Cisco IQ's Assets and Assessments REST APIs as MCP tools, so an AI assistant (e.g. Cursor, Claude Desktop) can query your entitled asset inventory, contracts, end-of-life lifecycle data, security advisories, and field notices directly.
[!WARNING] Cisco IQ APIs are in beta (public preview). Endpoint paths, request/response schemas, authentication, pagination, and error handling may change between releases without maintaining backward compatibility. Do not use this server for production integrations.
What this server does
It wraps the 16 documented Cisco IQ operations (as of the 2026-07-24 beta release,
API version 0.1.0) as read-only MCP tools:
| Resource | Tools |
|---|---|
| Assets | list_assets, get_asset, get_asset_lifecycle, get_asset_relationships, list_asset_security_advisories, list_asset_field_notices |
| Contracts | list_contracts, get_contract |
| Security Advisories | list_security_advisories, get_security_advisory, list_security_advisory_affected_assets, get_security_advisory_affected_asset |
| Field Notices | list_field_notices, get_field_notice, list_field_notice_affected_assets, get_field_notice_affected_asset |
All tools are GET-only; this server never performs write operations against Cisco IQ.
The server also transparently handles Cisco IQ's two-step auth flow: it exchanges your long-lived Personal Access Token (PAT) or Service Account Token (SAT) for a short-lived Bearer access token, caches it in memory, and refreshes it automatically before it expires — so tool calls never require you to think about tokens.
Prerequisites
- Node.js 18 or later
- A Cisco IQ account with permission to view the data you want to retrieve
- A Personal Access Token (PAT) or Service Account Token (SAT) (see below)
- Your Cisco IQ Account ID and Data Storage Region (
US,EMEA, orAPJC) — find both at Cisco IQ → Home → System Settings → Account Details
Generating a token
Personal Access Token (recommended for individual use)
- Log in to Cisco IQ.
- Click your name (top right) → User Settings.
- Under Personal Token Management, click Generate Token.
- Give it a name (e.g.
mcp-server), optionally a description, then click Generate Token. - Copy the token immediately — Cisco IQ will not show it again.
Service Account Token (for shared/automation use, Administrators only)
- Log in to Cisco IQ as an Administrator.
- Home → System Settings → Identity and Access → Add User.
- Select Service Account, name it, choose a role (Administrator or Viewer + resource groups), and save.
- Copy the generated token immediately — it is not shown again.
See Cisco's own Token Security Best Practices below before storing it.
Setup
npm install
npm run build
Copy .env.example to .env and fill in your values (this file is gitignored and
must never be committed):
cp .env.example .env
# Exactly one of these:
CISCO_IQ_PAT=your-personal-access-token
# CISCO_IQ_SAT=your-service-account-token
# Required for PAT auth; optional (but must match) for SAT auth
CISCO_IQ_ACCOUNT_ID=your-account-id
# Required: US, EMEA, or APJC
CISCO_IQ_REGION=APJC
Run it directly to confirm it starts:
npm start
You should see a line on stderr like:
[cisco-iq-mcp-server] Ready (region=APJC, auth=PAT). Cisco IQ APIs are beta/public preview - do not use for production integrations.
Using it from Cursor
Add an entry to your mcp.json (Cursor Settings → MCP, or ~/.cursor/mcp.json /
.cursor/mcp.json in a project). Do not hardcode the token value in this file if
it will be committed to a shared/synced location — prefer a local, gitignored config,
or reference an environment variable already set in your shell profile.
{
"mcpServers": {
"cisco-iq": {
"command": "node",
"args": ["/absolute/path/to/cisco-iq-mcp-server/dist/index.js"],
"env": {
"CISCO_IQ_PAT": "your-personal-access-token",
"CISCO_IQ_ACCOUNT_ID": "your-account-id",
"CISCO_IQ_REGION": "APJC"
}
}
}
}
For local development without building, you can instead run npm run dev
(tsx src/index.ts) as the command/args.
Sharing this with colleagues
This repo contains no credentials — everyone who uses it generates and supplies their own PAT/SAT (see Generating a token above). Never share your own token with a colleague; give them this repo instead:
- Clone it:
git clone https://github.com/rrizbaf/cisco-iq-mcp-server.git - Follow Setup to build it and create their own
.env(ormcp.jsonentry) with their own PAT/SAT, Account ID, and Region. - Each person's tool calls run under their own Cisco IQ identity and permissions — access to specific assets/contracts is governed by Cisco IQ itself, not this server.
Example tool calls
List up to 5 assets with critical/high security advisories:
{ "name": "list_assets", "arguments": { "hasCriticalOrHighSecurityAdvisories": true, "max": 5 } }
Get lifecycle milestones for a specific asset:
{ "name": "get_asset_lifecycle", "arguments": { "assetId": "85f9981e37312238b5c73020031a7b36", "milestoneType": "software" } }
List security advisories affecting a given asset:
{ "name": "list_asset_security_advisories", "arguments": { "assetId": "85f9981e37312238b5c73020031a7b36", "impact": ["Critical", "High"] } }
Pagination, filtering, and field selection
max(1-200, default 50) andoffsetcontrol page size/position on every collection tool.- Collection tool results include a
paginationobject ({ next?, prev? }) taken from Cisco IQ'sLinkresponse header — pass thenextURL'soffset/maxback in for the next page. Cisco IQ does not return a total result count. - Most list tools accept a
fieldsparameter (comma-separated) to request only the properties you need, which keeps responses small and LLM-context-friendly. - Array filters (e.g.
productFamily,serialNumber) accept multiple values.
Rate limits and error handling
Cisco IQ enforces both per-user and per-account rate limits:
| Scope | Requests/second | Requests/24h |
|---|---|---|
| User (PAT/SAT) | 10 | 5,000 |
| Cisco IQ Account | 25 | 25,000 |
This server automatically retries 502 Bad Gateway with bounded exponential backoff,
and 429 Too Many Requests by waiting for the shortest documented reset window (capped
at 30 seconds so a single tool call never blocks indefinitely). It does not retry
400, 401 (beyond one token refresh attempt), 403, 404, or 406 — those are
returned to the caller as a structured error (status, message, trackingId when
present) instead of being retried blindly.
Security notes
- Credentials live only in environment variables, read once at startup and held in memory. They are never logged, written to disk, or included in error messages.
.envis gitignored; only.env.example(with empty placeholders) is committed.- Short-lived access tokens are cached in memory only and refreshed automatically before they expire; they are never persisted.
- Per Cisco's own guidance:
- Store tokens in a secrets manager or other secure credential store.
- Do not put tokens in URLs, screenshots, log files, source code, or shared documents.
- Rotate tokens before they expire; revoke tokens immediately if exposed.
- Use the least-privileged role and resource group access required for the integration (prefer a Viewer-role SAT scoped to specific resource groups for read-only automation).
Project structure
src/
config.ts # env-var loading & validation (no hardcoded secrets)
auth.ts # TokenManager: PAT/SAT -> short-lived Bearer token
client.ts # CiqClient: query building, pagination, retry/backoff
errors.ts # CiqApiError + error-body parsing
types.ts # TS interfaces for documented response schemas
tools/
shared.ts # common Zod schemas & MCP result helpers
assets.ts # 6 asset-related tools
contracts.ts # 2 contract-related tools
securityAdvisories.ts # 4 security-advisory tools
fieldNotices.ts # 4 field-notice tools
index.ts # MCP server entrypoint (stdio transport)
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。