mcp-gitlab-workflow
An MCP server for issue-driven GitLab development, automating requirement analysis, issue creation, branching, coding, merge requests, and issue updates.
README
mcp-gitlab-workflow
mcp-gitlab-workflow is an MCP server for issue-driven GitLab development.
It can take a requirement or an existing GitLab issue and drive a standardized delivery flow: requirement analysis, issue creation, branching, coding, merge request creation, and issue updates. It also provides a set of atomic GitLab API tools for custom orchestration and fine-grained control.
1. Core Capabilities
workflow_*: higher-level tools that package common requirement-to-delivery flows into a single tool callgitlab_*: atomic GitLab API tools for custom orchestration and fine-grained control
Target projects for issues and code delivery are configured with WORKFLOW_ISSUE_PROJECT_ID and WORKFLOW_CODE_PROJECT_ID. See the environment variable section below for the full configuration matrix.
Workflow Tools
workflow_requirement_to_issue: analyze a requirement and create a GitLab issueworkflow_review_mr_post_comment: review a target merge request and post a review commentworkflow_issue_to_delivery: start from an existing issue and complete branch -> code change -> MR -> issue comment -> issue logworkflow_requirement_to_delivery: run the full end-to-end flow from a requirement
Atomic GitLab Tools
Issue tools fall back to WORKFLOW_ISSUE_PROJECT_ID by default, while code and merge request tools fall back to WORKFLOW_CODE_PROJECT_ID.
- Users and labels:
gitlab_get_current_user,gitlab_list_labels,gitlab_create_label,gitlab_update_label,gitlab_delete_label - Issues:
gitlab_create_issue,gitlab_get_issue,gitlab_get_issue_notes,gitlab_add_issue_comment,gitlab_get_issue_images - Repository:
gitlab_create_branch,gitlab_get_file,gitlab_commit_files,gitlab_upload_project_file - Merge requests:
gitlab_get_merge_request,gitlab_get_mr_notes,gitlab_create_merge_request,gitlab_create_mr_note,gitlab_get_mr_changes,gitlab_approve_mr,gitlab_unapprove_mr
2. Example
Using workflow_requirement_to_delivery as an example, it is best to explicitly name the tool in your prompt or wrap it in a higher-level custom instruction / skill so the LLM can reliably choose the intended workflow.
Input
Use workflow_requirement_to_delivery to add a forum interaction feature
Result
The workflow can automatically complete the flow below:
requirement analysis -> issue creation -> branch -> code change -> MR -> issue update

(1) Issue creation

(2) Code changes and MR submission based on the issue

(3) Issue comment

(4) Local issue-log.md

With this workflow, a developer can hand a requirement or an assigned issue directly to an agent and let it complete the full development flow in an issue-driven way.
3. Configuration
3.1 NPX
{
"mcpServers": {
"gitlab-workflow": {
"command": "npx",
"args": ["-y", "@chntif/mcp-gitlab-workflow"],
"env": {
"GITLAB_TOKEN": "YOUR_TOKEN",
"GITLAB_API_BASE_URL": "https://gitlab.com/api/v4",
"WORKFLOW_ISSUE_PROJECT_ID": "82346102",
"WORKFLOW_ISSUE_PROJECT_PATH": "tchen1690/test",
"WORKFLOW_CODE_PROJECT_ID": "82346102",
"WORKFLOW_CODE_PROJECT_PATH": "tchen1690/test",
"WORKFLOW_BASE_BRANCH": "develop",
"WORKFLOW_TARGET_BRANCH": "develop",
"WORKFLOW_LOCAL_REMOTE_NAME": "origin"
}
}
}
}
3.2 Codex
(1) Add from the terminal
codex mcp add gitlab-workflow \
--env GITLAB_TOKEN=YOUR_TOKEN \
--env GITLAB_API_BASE_URL=https://gitlab.com/api/v4 \
--env WORKFLOW_ISSUE_PROJECT_ID=80376102 \
--env WORKFLOW_ISSUE_PROJECT_PATH=tchen1690/test \
--env WORKFLOW_CODE_PROJECT_ID=80376102 \
--env WORKFLOW_CODE_PROJECT_PATH=tchen1690/test \
--env WORKFLOW_BASE_BRANCH=develop \
--env WORKFLOW_TARGET_BRANCH=develop \
--env WORKFLOW_LOCAL_REMOTE_NAME=origin \
-- npx -y @chntif/mcp-gitlab-workflow
(2) Or add it to config.toml
[mcp_servers.gitlab-workflow]
command = "npx"
args = ["-y", "@chntif/mcp-gitlab-workflow"]
[mcp_servers.gitlab-workflow.env]
GITLAB_TOKEN = "YOUR_TOKEN"
GITLAB_API_BASE_URL = "https://gitlab.com/api/v4"
WORKFLOW_ISSUE_PROJECT_ID = "80376102"
WORKFLOW_ISSUE_PROJECT_PATH = "tchen1690/test"
WORKFLOW_CODE_PROJECT_ID = "80376102"
WORKFLOW_CODE_PROJECT_PATH = "tchen1690/test"
WORKFLOW_BASE_BRANCH = "develop"
WORKFLOW_TARGET_BRANCH = "develop"
WORKFLOW_LOCAL_REMOTE_NAME = "origin"
3.3 Claude Code
claude mcp add gitlab-workflow \
-e GITLAB_TOKEN=YOUR_TOKEN \
-e GITLAB_API_BASE_URL=https://gitlab.com/api/v4 \
-e WORKFLOW_ISSUE_PROJECT_ID=80376102 \
-e WORKFLOW_ISSUE_PROJECT_PATH=tchen1690/test \
-e WORKFLOW_CODE_PROJECT_ID=80376102 \
-e WORKFLOW_CODE_PROJECT_PATH=tchen1690/test \
-e WORKFLOW_BASE_BRANCH=develop \
-e WORKFLOW_TARGET_BRANCH=develop \
-e WORKFLOW_LOCAL_REMOTE_NAME=origin \
-- npx -y @chntif/mcp-gitlab-workflow
You can also directly write the configuration in 3.1 into Claude Code's configuration file.
3.4 Local startup
{
"mcpServers": {
"gitlab-workflow": {
"command": "node",
"args": ["/gitlab-workflow-server/dist/src/server.js"],
"env": {
"GITLAB_TOKEN": "YOUR_TOKEN",
"GITLAB_API_BASE_URL": "https://gitlab.com/api/v4",
"WORKFLOW_ISSUE_PROJECT_ID": "82346102",
"WORKFLOW_ISSUE_PROJECT_PATH": "tchen1690/test",
"WORKFLOW_CODE_PROJECT_ID": "82346102",
"WORKFLOW_CODE_PROJECT_PATH": "tchen1690/test",
"WORKFLOW_BASE_BRANCH": "develop",
"WORKFLOW_TARGET_BRANCH": "develop",
"WORKFLOW_LOCAL_REMOTE_NAME": "origin"
}
}
}
}
3.5 About uv/uvx
This project is a Node.js package. npx is the recommended way to run it.
4. Environment Variables
Parameter resolution order:
tool args -> user-provided env -> built-in defaults
If the user does not explicitly pass a parameter such as project_id, the tool can use runtime configuration from environment variables. Some variables also have built-in defaults.
| Environment Variable | Purpose | Default | Required |
|---|---|---|---|
GITLAB_TOKEN |
GitLab API access token used by the server and all GitLab operations | None | Yes |
GITLAB_API_BASE_URL |
Base URL for the GitLab API | https://gitlab.com/api/v4 |
No |
WORKFLOW_ISSUE_PROJECT_ID |
Default target project ID for issue-related tools | None | No |
WORKFLOW_ISSUE_PROJECT_PATH |
Default issue project path used in templates and references | None | No |
WORKFLOW_CODE_PROJECT_ID |
Default target project ID for repository, branch, commit, and MR tools | None | No |
WORKFLOW_CODE_PROJECT_PATH |
Default code project path used in templates, logs, and output | None | No |
WORKFLOW_BASE_BRANCH |
Default base branch used before creating a new delivery branch | develop |
No |
WORKFLOW_TARGET_BRANCH |
Default merge request target branch | develop |
No |
WORKFLOW_LOCAL_REMOTE_NAME |
Default git remote used by local git workflows | origin |
No |
WORKFLOW_LABEL |
Default issue title prefix and fallback label | None | No |
WORKFLOW_ASSIGNEE_USERNAME |
Default assignee username for issues and merge requests | None | No |
WORKFLOW_ISSUE_LOG_PATH |
Path of the local issue log file | issue-log.md |
No |
WORKFLOW_UPDATE_ISSUE_LOG |
Whether delivery workflows update the local issue log by default | true |
No |
WORKFLOW_DELIVERY_METHOD |
Default delivery mode, supports local_git and remote_api |
local_git |
No |
WORKFLOW_CHECKOUT_LOCAL_BRANCH |
Whether remote_api delivery should sync and checkout the created branch locally |
false |
No |
WORKFLOW_LOCK_ISSUE_PROJECT_ID |
Restrict issue operations to a fixed issue project ID | None | No |
WORKFLOW_LOCK_CODE_PROJECT_ID |
Restrict code and MR operations to a fixed code project ID | None | No |
Recommended configuration
GITLAB_TOKEN,GITLAB_API_BASE_URL: required to connect to GitLabWORKFLOW_ISSUE_PROJECT_ID,WORKFLOW_CODE_PROJECT_ID: define the issue project and code project, and can point to the same projectWORKFLOW_ISSUE_PROJECT_PATH,WORKFLOW_CODE_PROJECT_PATH: improve rendered references and display output, but are optionalWORKFLOW_BASE_BRANCH,WORKFLOW_TARGET_BRANCH: should follow your team branching strategyWORKFLOW_LOCAL_REMOTE_NAME: usuallyorigin
If a specific operation should target values different from your default environment configuration, passing explicit tool arguments will override the configured values.
5. 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。