Nolane Habitat
Provides coding agents with a durable, revision-aware project workspace for semantic context, governed source changes, verification, task checkpoints, and observability through an MCP interface.
README
Nolane Habitat
Project intelligence for coding agents.
Nolane Habitat gives an agent a durable, revision-aware workspace for understanding and evolving a software project. It combines semantic project context, governed source changes, verification, task checkpoints, an MCP interface, and a read-only Observatory in one local system.
What Habitat brings to an agent
- Durable project memory — a SQLite-backed workspace records source identity, semantic evidence, context, checkpoints, and verification history.
- Focused semantic context — agents can start a task, inspect relevant objects, follow references, and continue with bounded context instead of rebuilding understanding from scratch.
- Governed source evolution — journaled mutations, source anchors, transaction recovery, and path-scoped verification support deliberate code changes.
- Live project awareness — Habitat reconciles source revisions and keeps task context aligned with the working project.
- Codex-native operation — an MCP adapter exposes task, context, inspection, change, verification, UI, checkpoint, and resume workflows to Codex.
- Clear observability — the loopback Observatory presents project and agent activity through a read-only runtime surface.
Quick start on Windows
Python 3.10+ is required. Clone the repository, then create a virtual environment with the development, MCP, and Python semantic capabilities:
git clone https://github.com/Nolane-x/Nolane-habitat.git
cd Nolane-habitat
python -m venv .venv
.\.venv\Scripts\python -m pip install -U pip
.\.venv\Scripts\python -m pip install -e ".[dev,mcp,python-semantic]"
Create a Habitat workspace beside the source project, then orient a task:
$source = (Resolve-Path .).Path
$workspace = "$source.habitat"
.\.venv\Scripts\habitat.exe create $source $workspace
.\.venv\Scripts\habitat.exe enter $workspace
.\.venv\Scripts\habitat.exe orient $workspace "map the authentication flow"
The workspace is intentionally separate from the source directory, keeping Habitat state independent from the project it understands.
Use it with Codex
Register the initialized workspace as an MCP server:
$python = (Resolve-Path .\.venv\Scripts\python.exe).Path
codex mcp add nolane-habitat -- $python -m habitat.mcp_adapter $workspace --no-open-observatory
codex mcp list
Install the bundled Codex skills from the local checkout:
$repo = (Resolve-Path .).Path
codex plugin marketplace add $repo
codex plugin add nolane-habitat@personal
Start a Codex task with $nolane-habitat to use the project workspace, or $nolane-habitat-maintainer when improving Habitat itself. The full setup and day-to-day workflow are in Codex integration.
A practical development loop
- Create or open a Habitat workspace for the project.
- Start a task and gather bounded semantic context.
- Inspect symbols and references before editing.
- Apply an authorized source change through the governed workflow.
- Verify affected paths and checkpoint the result for the next agent.
This loop makes context, changes, evidence, and handoffs durable across longer-running agent work.
Core surfaces
| Surface | What it provides |
|---|---|
| CLI | Workspace creation, orientation, inspection, mutation, verification, checkpoints, and resume. |
| MCP | Codex tools for task context, source understanding, governed changes, verification, and handoff. |
| Semantic providers | Python and TypeScript project evidence with source anchors and relationship data. |
| Mutation engine | Journaled source writes, rollback support, and recovery-aware transaction handling. |
| Observatory | A loopback, read-only view of project and agent activity. |
Documentation
- Install and run Habitat
- Connect Habitat to Codex
- Use the agent protocol
- Explore the architecture
- Review the changelog
Verify a checkout
.\.venv\Scripts\python tools\run_test_matrix.py --workers 1 --timeout 180
Nolane Habitat 0.1.0-alpha.18 is ready for local project cognition, governed agent workflows, and Codex integration.
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。