Linear MCP
Custom MCP server for Linear that enables managing issues, projects, and workspaces through natural language, backed by raw GraphQL.
README
Linear MCP
Custom Bun/TypeScript MCP server for Linear, backed by raw GraphQL and used through the local MCP control plane.
- Active local source:
/Users/jonas/.agents/mcp/servers/linear - Workspaces:
biz,personal, andjonas-test-workspace(testalias) - Plan levels:
bizandpersonalare Basic;jonas-test-workspaceis Free. - Auth:
LINEAR_BIZ_TOKEN,LINEAR_PERSONAL_TOKEN, andLINEAR_TEST_WORKSPACE_TOKEN - Generated tool reference: CAPABILITIES.md
Agent Usage
Fresh sessions should use one source-of-truth chain:
- Runtime MCP tool discovery for callable names, descriptions, and schemas.
- CAPABILITIES.md for the generated human-readable index, examples, and tool-choice guidance.
/Users/jonas/.agents/skills/linear/SKILL.mdfor workspace routing and operating policy.- Live MCP reads for IDs and current Linear workspace state.
Do not copy tool tables into skills, README sections, or Linear comments. If tool behavior changes, update the implementing ToolDef metadata/examples under tools/, regenerate CAPABILITIES.md, and keep the skill focused on policy rather than a second reference.
Setup
bun install
cp .env.example .env
bun run start
The local production launch path uses /Users/jonas/.agents/mcp/wrappers/linear.sh, which loads tokens through the neutral MCP wrapper layer. The smoke scripts in package.json assume that local control-plane layout.
Scripts
bun run prepare:repo
bun run verify
bun run smoke:tools
bun run smoke:tools:local
bun run smoke:views
bun run smoke:live -- --workspace jonas-test-workspace --scenario comments
bun run smoke:live -- --workspace biz --scenario discord --discord-url https://discord.com/channels/<guild>/<channel>/<message>
bun run seed:test-workspace
Use bun run prepare:repo before committing source changes. It refreshes the generated capabilities reference, then runs the normal local verification path. GitHub CI runs docs, build, and credential-free tool discovery; live Linear calls still depend on local credentials and the local MCP wrapper layout.
bun run smoke:live is opt-in and mutates Linear. It defaults to jonas-test-workspace and also accepts --workspace personal or --workspace biz. It creates durable sandbox anchors named Linear MCP Sandbox, creates disposable fixtures named MCP Smoke <domain> <timestamp>, tracks created IDs, and cleans current-run fixtures in finally. Supported scenarios are favorites, labels, duplicate, organize, comments, views, icons, notifications, subscriptions, templates, discord, and all. The discord scenario is intentionally not part of all because it needs --discord-url and the biz Discord integration. The script warns about stale MCP Smoke fixtures before it starts; review those warnings separately from current-run cleanup results.
bun run seed:test-workspace is also mutating and intentionally durable. It targets jonas-test-workspace by default, creates a fleshed-out MCP capability lab, and writes a local report under reports/.
Layout
index.ts
workspaces.ts
client.ts
tools/
registry.ts
issues.ts
projects.ts
initiatives.ts
views.ts
templates.ts
files.ts
attachments.ts
...
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。