workiq-proxy
Enhances Microsoft Work IQ with broader MCP client compatibility, structured search tools, an interactive REPL, and an OpenAI-compatible HTTP API.
README
workiq-proxy
Extends Microsoft Work IQ with broad MCP client support, structured search tools, an interactive REPL, and an OpenAI-compatible HTTP API.
What it does
workiq-proxy enhances Work IQ's MCP server so it works seamlessly with Claude Code, VS Code, Codex, and other MCP clients.
In MCP mode the proxy sits between the MCP client and the Work IQ server:
MCP Client <-stdio-> workiq-proxy <-stdio-> workiq mcp
It:
- Extends MCP capabilities — Adds support for
prompts/list,resources/list, andresources/templates/listso every MCP client connects cleanly - Advertises full capabilities — Ensures clients see a fully capable server during initialization
- Future-proof — Automatically steps aside when Work IQ gains these capabilities natively
- Error enrichment — Wraps opaque errors with troubleshooting context (EULA not accepted, token protection / error 530084, Entra ID conditional access / AADSTS, interactive login required, and "Failed to create conversation")
- Synthetic tools — Exposes 7 domain-specific tools (
search_emails,search_documents,search_chats,search_channels,search_meetings,search_people,search_external) backed byask_work_iq - Interactive REPL — When run from a terminal with no arguments, launches a Bubble Tea UI with slash commands (
/ask,/emails,/docs,/chats,/channels,/meetings,/people,/accept-eula,/tools,/help,/quit), session history, and glamour markdown rendering - OpenAI-compatible HTTP API —
workiq-proxy servestarts a local server with/v1/chat/completions(streaming and non-streaming) and/v1/models - CLI passthrough — When run from an interactive terminal with arguments, delegates to the underlying
workiqCLI (EULA acceptance, auth, queries) so you only need one package - Traffic logging — Logs all JSON-RPC messages to
~/.work-iq-cli/mcp-traffic.log
Install
curl -fsSL https://stuffbucket.github.io/workiq-proxy/install.sh | sh
The installer detects (or installs) Node.js and sets up workiq-proxy via npm.
On Windows (PowerShell):
irm https://stuffbucket.github.io/workiq-proxy/install.ps1 | iex
Alternatively, install directly with npm:
npm install -g @stuffbucket/workiq-proxy
This also installs @microsoft/workiq as a dependency — no separate setup needed. The npm package provides both workiq-proxy and wiq as binaries.
Or download a binary from GitHub Releases and place it in your PATH.
Prerequisites
Before using the proxy, set up Work IQ from your terminal:
# First run: accepts EULA interactively, then triggers browser auth
npx -y @stuffbucket/workiq-proxy ask
The interactive session walks you through EULA acceptance and signs you in via the browser. Once complete, the MCP server reuses cached tokens silently.
You can also accept the EULA directly:
npx -y @stuffbucket/workiq-proxy accept-eula
To verify auth works non-interactively:
npx -y @stuffbucket/workiq-proxy ask -q "What's on my calendar?"
Commands
| Command | Description |
|---|---|
| (no args, TTY) | Start the interactive REPL |
ask -q "..." |
Ask Microsoft 365 Copilot a question (passthrough to workiq) |
accept-eula |
Accept the Work IQ EULA via the MCP accept_eula tool |
mcp |
Start MCP proxy server (stdio) — also the implicit mode when no TTY |
serve |
Start OpenAI-compatible HTTP API on localhost |
json |
Interactive JSON-RPC testing mode |
version |
Show workiq CLI version (passthrough) |
help |
Show usage summary |
Configure
Claude Code
Project-level (adds to .mcp.json in current directory):
claude mcp add --transport stdio workiq --scope project -- npx -y @stuffbucket/workiq-proxy
User-level (available in all projects):
claude mcp add --transport stdio workiq --scope user -- npx -y @stuffbucket/workiq-proxy
Or add to .mcp.json manually:
{
"mcpServers": {
"workiq": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@stuffbucket/workiq-proxy"]
}
}
}
VS Code / GitHub Copilot
Project-level — add to .vscode/mcp.json:
{
"servers": {
"workiq": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@stuffbucket/workiq-proxy"]
}
}
}
User-level — add to VS Code user settings (settings.json):
{
"mcp": {
"servers": {
"workiq": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@stuffbucket/workiq-proxy"]
}
}
}
}
OpenAI Codex CLI
Add to .codex/mcp.json:
{
"mcpServers": {
"workiq": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@stuffbucket/workiq-proxy"]
}
}
}
Flags
| Flag | Default | Description |
|---|---|---|
--workiq-cmd |
npx -y @microsoft/workiq |
Base command to reach the Work IQ CLI (proxy appends mcp for MCP mode) |
--log-file |
~/.work-iq-cli/mcp-traffic.log |
Override log file path |
--no-log |
false |
Disable traffic logging |
--port |
11435 |
Port for serve mode (seeks next available if busy) |
Example with a globally installed Work IQ CLI:
{
"mcpServers": {
"workiq": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@stuffbucket/workiq-proxy", "--workiq-cmd", "workiq"]
}
}
}
Verify
After configuring, check that the server connects:
- Claude Code: Run
/mcpto see server status and available tools - VS Code: Check the MCP server status in the output panel
You should see 9 tools: accept_eula, ask_work_iq, plus the 7 synthetic search tools.
Troubleshoot
The proxy enriches five categories of upstream errors with actionable hints:
| Error pattern | Likely cause | Fix |
|---|---|---|
| EULA / license agreement | Work IQ EULA not accepted | Run workiq-proxy accept-eula |
| Error 530084 / token protection | Org requires token protection CA policy | Ask IT admin to exempt the Work IQ CLI app |
| AADSTS / security policy | Entra ID conditional access policy blocking | Check the error code and contact IT admin |
| InteractionRequired | Browser login needed | Run workiq-proxy ask -q "What's on my calendar?" to re-authenticate |
| Failed to create conversation | Auth token expired or Copilot license inactive | Run workiq-proxy ask -q "What's on my calendar?" to re-authenticate |
Server not connecting — Check ~/.work-iq-cli/mcp-traffic.log for the full JSON-RPC transcript.
Development
# Build
make build
# Test
make test
# Lint (Go + JS)
make lint
# Cross-compile all platforms
make all
Programmatic API
Use workiq-proxy as a dependency to embed Work IQ in your own Node.js project:
npm install @stuffbucket/workiq-proxy
Pipe-based (no port, no network)
const { createClient } = require("@stuffbucket/workiq-proxy");
const client = await createClient();
const answer = await client.ask("What's on my calendar?");
console.log(answer);
const emails = await client.searchEmails({ from: "alice", date_range: "last week" });
const docs = await client.searchDocuments({ keywords: "quarterly report" });
const chats = await client.searchChats({ person: "bob", keywords: "standup" });
await client.close();
createClient() spawns workiq-proxy over stdio pipes — no port, no HTTP. The MCP handshake happens automatically.
| Method | Description |
|---|---|
ask(question) |
Ask Microsoft 365 Copilot a question |
searchEmails(params) |
Search emails (from, subject, keywords, date_range) |
searchDocuments(params) |
Search documents (filename, keywords, site, file_type) |
searchChats(params) |
Search chats (person, keywords, date_range, channel) |
callTool(name, args) |
Call any MCP tool by name |
listTools() |
List all available tools |
close() |
Shut down the child process |
Options: createClient({ workiqCmd?, noLog? })
HTTP server (OpenAI-compatible)
If you need an HTTP API (e.g. for non-Node consumers):
const { createServer } = require("@stuffbucket/workiq-proxy");
const server = await createServer(); // starts serve --json
const models = await server.listModels(); // /v1/models
const reply = await server.chat("What's on my calendar?");
console.log(reply.choices[0].message.content);
await server.close();
| Method | Description |
|---|---|
server.url |
Base URL (e.g. http://127.0.0.1:11435) |
server.listModels() |
List available models |
server.chat(message) |
Send a chat message (string or full OpenAI request body) |
server.chat(body, { stream: true }) |
Stream SSE chunks |
server.onEvent(fn) |
Subscribe to JSONL log events; returns unsubscribe function |
server.events |
Array of all captured log events |
server.close() |
Shut down the server |
Options: createServer({ port?, workiqCmd?, noLog? })
License
MIT — see LICENSE. Charmbracelet libraries are used under the MIT license (see third-party notices in LICENSE).
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。