Nexul
One loop from docs to deploy, on servers you own.
接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令? 兩種安裝方式的差異
[!WARNING] Nexul is in beta. Expect rough edges and breaking changes between releases. Proceed with caution before trusting it with production work.
Nexul is a self-hosted platform for teams that ship software with coding agents in the loop. Docs, tickets, chat, CI runners, and deploys to your own servers live in one place, and one MCP server exposes all of it, so an agent works through the same tools people do.
One install covers what these do, area by area:
Phew, that is a lot of tabs to close.
Three repositories shaped Nexul.
mattpocock/skills showed how short the path from an idea to a set of scoped tickets can be: grill the idea against the docs, map it, and the tickets fall out. A developer gets from "I think we should" to a board of work in one sitting.
pingdotgg/t3code showed how natural it is to run several pieces of work in parallel once an agent does the typing, and how much that changes the way code should be written.
pstack, Lauren Tan's engineering-discipline pack, showed how far short principle skills and playbooks go when an agent checks its own work against them.
Nexul takes all three ideas and applies them to the whole software delivery loop, from the doc to the deploy, on your own servers.
Every piece of work is linked and searchable in one place: docs, tickets, code, reviews, deploys, server topology, and run logs. One MCP server exposes all of it, so an agent can run the whole loop without leaving the system.
-> Docs to land ideas; not clear? run /wayfinder play or /grill-with-docs to clear it
-> Tickets derived from docs (isolated, scoped)
-> Work on tickets (human or LLM agent, owner's choice)
-> Code + Pull Requests (linked back to ticket + doc)
-> Code review (linked, indexed)
-> Deploy (runner executes, topology updates live)
-> Index everything -> the next loop starts with full context
Every step is a use case. The web app reaches the use-case layer through an HTTP gateway, and MCP exposes its registered tools through the same layer. A personal access token carries the user's permissions, and calls made through it are attributed to that user. The adapters are separate surfaces over one source of behavior.
Everything below is in the AGPL edition and runs on your own servers. The guide covers each area.
Go (version pinned in go.mod) and Bun are the only prerequisites.
CONTRIBUTING.md is the short version, and the
contributor docs have the rest.
Coding agents are contributors too: AGENTS.md is written for
them, and it is a good read for people as well. ROADMAP.md
says where the product is going, .scratch/ holds the open
specs and tickets, and CONTEXT.md is the vocabulary.
Nexul is open core: one repository, two licenses.
ee/ is AGPL-3.0-or-later. Anyone can run it, modify
it, and self-host it at any company size, as long as they follow the AGPL
and publish their changes if they offer a modified copy over a network.ee/ is under a commercial license, see ee/LICENSE.
The code is public so you can read and patch it, but running it in
production needs a Nexul subscription. Today that directory holds only the
license.Outside contributions need a Contributor License Agreement so the project can
keep licensing both parts. The text is in CLA.md; you agree to
it by ticking the box in the pull request template.
What the subscription is meant to include.
Self-hosted platform for project management, docs, CI runners, and deployment to your own servers, with a first-class MCP server so coding agents work through the same tools people do.
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Nexul 透過 stdio 在本機執行。
上次檢查時間:2026-11-15。 Self-hosted platform for project management, docs, CI runners, and deployment to your own servers, with a first-class MCP server so coding agents work through the same tools people do.