Gridlace
Enables deterministic dependency reasoning over declared graphs, allowing users and AI agents to validate acyclicity, resolve targets, identify runnable parallel layers, assess impact, and explain blockers.
README
Gridlace
A Deterministic Dependency Engine for humans and AI agents.
The name evokes points laced into a changing pattern: move one declared relation and the consequences across the structure change. It is an image, not a claim that the engine uses a regular grid or provides a graph editor.

It answers five questions over relations you declare:
- Is this dependency graph valid and acyclic?
- What remains before these targets can be completed?
- What can run now and in which parallel layers?
- What will these changes affect?
- Why is this target blocked or affected?
The engine proves derivations over declared relations. It does not prove that the declared business relations are true.
This repository is deliberately limited to dependency reasoning. It is not a general orchestration standard, provider-selection framework, task runner, graph discovery system, graph database, or visual graph editor.
Graph contract
{
"schema": "agent-deps/v1",
"nodes": [
{ "id": "schema" },
{ "id": "backend" },
{ "id": "release" }
],
"requires": [
{ "dependent": "backend", "prerequisite": "schema" },
{ "dependent": "release", "prerequisite": "backend" }
]
}
Public edges never use ambiguous from / to keys.
Develop
Use Node.js 22 or newer.
This repository is the public source distribution. It is not currently published to the npm registry.
npm ci
npm run check
npm run start:ui
After npm run check, CLI requests can be run from a clone as operation-specific
JSON read from standard input or an explicit --input file:
node dist/adapters/cli.js resolve --input request.json --pretty
If you install a locally packed copy, the same entry point is available as
dependency-engine.
The MCP server exposes graph_validate, dependency_resolve,
dependency_impact, dependency_slice, and dependency_explain. The local
HTTP adapter exposes matching POST /v1/* routes and GET /health. When the
npm package is installed from a local pack rather than cloned, run the MCP entry point as
node node_modules/@openadam/dependency-engine/dist/adapters/mcp.js.
Start the built local HTTP adapter with:
node dist/adapters/http-server.js --host 127.0.0.1 --port 8787
The HTTP adapter defaults to 127.0.0.1 and is intended for trusted local use.
It does not provide authentication, authorization, or TLS. Do not expose it
directly to an untrusted network; see the security policy.
License
Copyright 2026 openAdam. Licensed under the Apache License 2.0. Bundled browser dependencies are documented in the third-party notices. The license does not grant rights to the Gridlace or openAdam names; see the brand policy.
See the product identity, product model, and review contract for exact boundaries.
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。