DevRPG-MCP
An MCP server that turns your development workflow into an RPG with persistent memory, tracking quests, perks, and skills across projects.
README
DevRPG-MCP
An MCP server that turns your development workflow into an RPG with persistent memory across projects.
Status: 0.1.0-alpha — initial bootstrap. API and storage schema are unstable until 1.0.0.
What is this?
DevRPG-MCP is a Model Context Protocol server that adds a long-term memory and progression layer to your work with LLM-coding assistants (Claude Code, Cursor, any MCP client). It tracks Quests, Subquests and Tasks; unlocks Perks for capabilities you actually use; stores Resources (domains, VPSes, code templates) and Skills as a layer above your existing skill files.
It is intentionally not a productivity tracker. It is a narrative engine — the agent inside your IDE turns work into a Hero's Journey arc, with a Game Master that observes raw transcripts and applies regulated XP / Karma / Divine Insight rules.
Quick start
# install (private repo for now — public after MVP)
pip install git+https://github.com/Zver1013/devrpg-mcp@main
# connect to Claude Code
claude mcp add --scope user devrpg-mcp -- devrpg-mcp
# initialize in current project
devrpg init .
After this, the next time you open Claude Code in a project, the DevRPG tools are available — the agent will use them according to regulations in system/regulations/.
Concept
Three roles:
- Operator — you, the human developer
- Agent — the LLM (Claude / GPT / etc.) acting as Game Master
- System — a deterministic sub-agent that audits work and applies XP / Karma rules from raw transcripts
See docs/concepts/system.md for the full philosophy and docs/design/ for technical specs (DB schema, MCP tools API, code-vs-LLM boundary).
Tools API
25 MCP tools across 8 categories — character, quest, subquest, task, perk, resource, skill, system, plus universal inventory_check. 24 of them are deterministic Python; one (system_snapshot) spawns a sub-agent for semantic evaluation.
Full reference: docs/design/mcp-tools-api.md.
Configuration
Per-project configuration lives in system/config.md (which modules are active, which narrative philosophy applies). Hooks for Claude Code are in templates/hooks.settings.json.
License
MIT — see LICENSE.
Author
Nazar Alekseev · zver1013@gmail.com
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。