mcp-mantis
Provides AI assistants with direct access to MantisBT issue trackers via the REST API, enabling reading, updating, resolving, and semantically searching issues.
README
mcp-mantis
An MCP stdio server that gives an AI assistant direct access to a MantisBT tracker, oriented to issue resolvers: read issues (with images), change status, resolve, assign, comment and link — plus local semantic search over issues.
REST, not SOAP. This server uses the MantisBT REST API (Mantis 2.27+), the supported and non-deprecated integration path. The legacy SOAP API is not used and does not need to be enabled.
Setup
Requires Node.js 22+ (uses the built-in global fetch).
npm install
cp .env.example .env
Then edit .env:
| Variable | Required | Notes |
|---|---|---|
MANTIS_REST_BASE_URL |
yes | Must include index.php when URL rewriting is disabled on the host. |
MANTIS_TOKEN |
yes | API token from Mantis: My Account → API Tokens → Create. No username/password; the token is the Authorization header. |
MANTIS_READONLY |
no | false (default) enables write tools. Set true for read-only. A blank value keeps the default. |
MANTIS_SEARCH_ENABLED |
no | true (default) registers semantic search tools. Needs the optional @huggingface/transformers. |
MANTIS_SEARCH_MODEL / MANTIS_SEARCH_DIR / MANTIS_SEARCH_THREADS |
no | Embedding model, index directory, ONNX threads. |
MANTIS_TIMEOUT_MS, MANTIS_METADATA_TTL_MS, MANTIS_DEFAULT_PAGE_SIZE, MANTIS_MAX_PAGE_SIZE |
no | Tuning. |
Register with Claude Code
From npm (published package), no local checkout needed — the token is supplied here, never stored in the repo:
claude mcp add mantis \
-e MANTIS_REST_BASE_URL=https://mantis.example.com/api/rest/index.php \
-e MANTIS_TOKEN=your-token-here \
-- npx -y @grec0/mcp-mantis
Or from a local build:
npm run build
claude mcp add mantis -- node /path/to/mcp-mantis/dist/mcp.js
Equivalent .mcp.json (the env block is where MANTIS_TOKEN belongs):
{
"mcpServers": {
"mantis": {
"command": "npx",
"args": ["-y", "@grec0/mcp-mantis"],
"env": {
"MANTIS_REST_BASE_URL": "https://mantis.example.com/api/rest/index.php",
"MANTIS_TOKEN": "your-token-here",
"MANTIS_READONLY": "false"
}
}
}
}
Tools
Read
| Tool | Purpose |
|---|---|
mantis_whoami |
Authenticated user + accessible projects. |
mantis_list_projects / mantis_get_project |
Projects with versions, categories, custom fields. |
mantis_list_project_users |
Project members (valid handler names for assignment). |
mantis_list_issues |
Paginated issues, lightweight summary by default (use mantis_get_issue for full detail). Client-side filters: status (name or "open"), assigned_to, reporter_id, date range. |
mantis_get_issue / mantis_get_issues |
One issue (complete) / batch fetch by ids. |
mantis_get_issue_attachment |
Download an attachment; images come back viewable. |
mantis_get_enums / mantis_get_config |
Enumerations ({id,name,label}) / raw config options. |
mantis_list_filters |
Stored filters (feed a filter_id into mantis_list_issues). |
Resolve / write (registered when MANTIS_READONLY=false)
| Tool | Purpose |
|---|---|
mantis_update_issue |
Partial update: status, resolution, handler, priority, fields… Accepts localized/canonical enum names. Optional note and dry_run. |
mantis_resolve_issue |
Sets both status and resolution (defaults resuelta / fixed) + optional note. |
mantis_assign_issue |
Assign to a handler by login name (or id), optional status change + note. |
mantis_add_note |
Add a public/private note. |
mantis_add_relationship / mantis_remove_relationship |
Link issues (related, duplicate, blocks, depends…). |
Semantic search (registered when MANTIS_SEARCH_ENABLED=true)
| Tool | Purpose |
|---|---|
mantis_rebuild_search_index |
Build/update the local vector index from issues (first run downloads the model). |
mantis_search_issues |
Natural-language search (multilingual). Optional select, highlight, date filters. |
mantis_search_index_status |
Index fill level and last sync. |
Enum names accept the instance's localized names and the canonical English labels (e.g. asignada
or confirmed). mantis_list_issues status="open" uses the live bug_resolved_status_threshold.
Ad-hoc free-text search is not a server-side REST filter — use mantis_search_issues or a stored filter_id.
Commands
| Command | Purpose |
|---|---|
npm run dev:mcp / npm run start:mcp |
Run the MCP stdio server (reload / built). |
npm run build |
Compile TypeScript into dist/. |
npm run check |
Type-check, lint and format-check. |
npm run dev |
Express health server (GET /health, GET /health/mantis). |
Project structure
src/
├── config.ts # Validated environment configuration
├── constants.ts # Enum aliases, relationship types
├── date-filter.ts # Shared date-range filtering
├── mantis.ts # MantisBT REST client (fetch, enum/handler resolution, caching)
├── tools/ # meta, issues, notes, relationships, shared helpers
├── search/ # embedder (Transformers.js), vector store, sync, search tools
├── server.ts # Tool registration
├── mcp.ts # MCP stdio executable
└── index.ts # Express health endpoints
Security
.env is git-ignored (only .env.example is committed) — never commit the token. Write access is
scoped to the token owner's Mantis permissions; set MANTIS_READONLY=true to disable all mutations.
Health endpoints expose no credentials or user data. Run npm run check before committing.
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。