#
ENJA

AAOSP

MCP Server

rufolangus/AAOSP

AAOSP is an experimental Android 15 fork that natively integrates the Model Context Protocol (MCP) and an on-device LLM to enable agentic OS behaviors and secure tool-calling.

24 5No licenseUpdated 2026-04-23
View on GitHubOpen source file

Overview

AAOSP modifies Android 15 to treat AI capabilities as core operating system services. It embeds an on-device Qwen model inside system_server through a llama.cpp JNI bridge, exposes MCP server declarations directly in AndroidManifest.xml files, manages tool execution via a binder-based agentic loop, and enforces HITL (human-in-the-loop) user consent for sensitive tool calls.

Capabilities

  • ▸Native MCP server declarations via AndroidManifest.xml parser
  • ▸On-device LLM execution inside system_server via LlmManagerService
  • ▸Binder-based tool call dispatch and response round-tripping
  • ▸Human-in-the-loop consent gates for restricted MCP tools
  • ▸Custom Android build configuration and Cuttlefish virtual device runner

Best for

Building and testing custom Android system software with native LLM integration, Developing agentic applications that can securely execute on-device MCP tools, Experimenting with OS-level AI runtime management, privacy consent gates, and binder-based tool dispatch

Works with

Claude CodeCodex
Evoa Score breakdown= Σ (score × weight)
59
Task usefulness20%88 → +17.6

実タスクにどれだけ役立つか（機能の豊富さ・用途の明確さ）。 — AIによるcapabilities/use-cases解析

Code quality15%90 → +13.5

実装・指示の品質。 — AIによるSKILL.md/README解析

Maintenance15%70 → +10.5

リポジトリがどれだけ活発に保守されているか。 — GitHub 最終push日時の新しさ

Documentation12%45 → +5.4

ドキュメントの充実度・分かりやすさ。 — README/独自要約の情報量

Security15%50 → +7.5

危険・不審な挙動が無いか。 — AIによるセキュリティレビュー

Originality10%95 → +9.5

ありふれたラッパーではない独自性。 — AIによる独自性判定

Popularity8%39 → +3.1

コミュニティの採用度。 — GitHub Stars/Forks（対数スケール）

Compatibility5%50 → +2.5

対応AIエージェントの広さ。 — AIによる対応エージェント判定

Weighted total69.6 / 100

ライセンス不明/制限あり（Red）のSkillは総合スコアに0.85倍の補正を適用します。 ランキングはこのScoreのみで決まり、広告で変わりません。 算出方法の詳細 →

Security considerations

Includes a HITL (Human-in-the-Loop) consent store and consent gate for actions requiring explicit user permission, alongside SELinux policy overlays.

Categories

CodingAI EngineeringDevOpsAutomation

Summary and analysis are original content generated by AI Skills Rank. The skill's source text is not reproduced here — view it on the linked repository.