A
licenseNot graded
qualityC
maintenanceEnables AI agents to discover, install, and manage SKILL.md skills from a Git-backed registry via MCP tools for search, install, and list operations.
20 npm
2
MIT
agy-mcp
by cyeezy08
Allows installing skills from GitHub repositories by cloning a repo, extracting its SKILL.md, and installing it as a skill.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@agy-mcp List my installed Antigravity skills"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
agy-mcp
MCP server for Antigravity IDE — skill management, config bridge, and agent orchestration.
Wraps Antigravity's local ~/.agents/ directory (skills, .skill-lock.json, config.json) as MCP tools, so an assistant can list, read, install, and search Antigravity skills, and read or write Antigravity configuration, without shelling out manually.
Status (0.2.0)
Single-file server (src/agy_mcp/server.py) exposing 11 tools plus agy_help, covered by a 34-test pytest suite (skill CRUD, config, GitHub install with stubbed clone, security regressions) running in CI on Linux + Windows, Python 3.11 and 3.12.
Security note: 0.1.0 joined caller-supplied skill names directly into filesystem paths, so agy_delete_skill("..") could delete arbitrary directories and agy_read_skill("../victim") could read files outside the skills directory. 0.2.0 validates every skill name against a strict charset and containment-checks the resolved path before any read, write, or delete. Because this server is driven by an LLM, treat these regressions as load-bearing — see tests/test_security.py before weakening name validation.
Related MCP server: OSource-Manager
Tools
Tool | Description |
| List installed skills from |
| Read a skill's full |
| Search skill names and content |
| Write or update a skill's |
| Remove an installed skill |
| Clone a GitHub repo, pull a |
| Read/write Antigravity's |
| List configured LLM providers from config |
| Read |
| Report the installed Antigravity CLI version |
| List all tools |
Quick start
cd D:\Dev\repos\agy-mcp
uv sync
uv run agy-mcpAdd to claude_desktop_config.json:
{
"mcpServers": {
"agy-mcp": {
"command": "uv",
"args": ["run", "--directory", "D:\\Dev\\repos\\agy-mcp", "agy-mcp"]
}
}
}By default it reads and writes ~/.agents/ — override with the AGENTS_DIR environment variable (resolved per call, so it can also be set just for tests).
Development
uv sync --group dev # pytest, pytest-asyncio, ruff, pre-commit
uv run pytest -q # 34 tests, < 1s, no network
uv run ruff check src tests
uv run ruff format src tests
just bootstrap # Windows: also installs the pre-commit hookCI (.github/workflows/ci.yml) runs lint + tests on ubuntu-latest and windows-latest across Python 3.11/3.12; scripts/ci.ps1 mirrors the same gates locally on Windows.
Requirements
Python 3.11+
FastMCP 3.4+
Available Tools
12 toolsagy_delete_skillA
Permanently remove an installed skill.
Return Format
{"success": bool, "name": str}
Examples
agy_delete_skill("old-skill")
Inputs: skill_name (required), ctx. Returns: object
agy_get_configB
Read the full Antigravity configuration.
Return Format
{"config": dict, "path": str}
Inputs: ctx. Returns: object
agy_helpB
Show all available agy-mcp tools and usage.
Inputs: ctx. Returns: object
agy_install_skillB
Create or update a skill by writing its SKILL.md.
Also registers the skill in .skill-lock.json as a local skill.
Return Format
{"success": bool, "name": str}
Examples
agy_install_skill("my-skill", "# My Skill\n\nBe helpful and concise.")
Inputs: content (required), skill_name (required), ctx. Returns: object
agy_install_skill_from_githubA
Install a skill directly from a GitHub repository.
Clones the repo temporarily and extracts the skill directory.
Return Format
{"success": bool, "repo": str, "skill_name": str}
Examples
agy_install_skill_from_github("sandraschi/python-expert") agy_install_skill_from_github("sandraschi/mcp-central-docs", "skills/fleet-expert")
Inputs: repo (required), ctx, skill_path. Returns: object
agy_list_providersA
List configured LLM providers from Antigravity config.
Reads the providers section from config.json.
Return Format
{"providers": [{"name": str}], "count": int}
Inputs: ctx. Returns: object
agy_list_skillsA
List all installed Antigravity agent skills with source info.
Scans ~/.agents/skills/ and merges metadata from .skill-lock.json.
Return Format
{"skills": [{"name": str, "source": str, "source_type": str}], "count": int}
Inputs: ctx. Returns: object
agy_read_skillA
Read a skill's full SKILL.md content.
Return Format
{"name": str, "content": str, "meta": dict}
Examples
agy_read_skill("python-expert")
Inputs: skill_name (required), ctx. Returns: object
agy_search_skillsB
Search across all installed skill names and content.
Return Format
{"results": [{"name": str, "match": str}], "count": int}
Examples
agy_search_skills("python") agy_search_skills("docker")
Inputs: query (required), ctx. Returns: object
agy_set_configA
Set a configuration value in Antigravity config.json.
Creates the config file if it doesn't exist.
Return Format
{"success": bool, "key": str, "value": any}
Examples
agy_set_config("default_model", "claude-sonnet-4-6")
Inputs: key (required), value (required), ctx. Returns: object
agy_skill_lock_statusB
Read the .skill-lock.json to see installed skill sources and versions.
Return Format
{"version": int, "skills": [{"name": str, "source": str, "source_type": str}], "count": int}
Inputs: ctx. Returns: object
agy_versionA
Get the installed Antigravity CLI version.
Return Format
{"version": str, "path": str}
Inputs: ctx. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
A3.7/5.0
Scored across 12 tools
Disambiguation4/5
Most tools target clearly distinct actions (list/read/search/install/delete skills vs. config get/set), but agy_list_skills and agy_skill_lock_status both report installed skill source info and could be confused. The detailed return formats and examples help resolve most ambiguity.
Naming Consistency4/5
Tools consistently use the agy_ prefix with snake_case and a mostly verb_noun pattern like agy_list_skills and agy_set_config. Minor exceptions like agy_help, agy_version, and agy_skill_lock_status deviate from the verb_noun convention but remain readable and predictable.
Tool Count5/5
12 tools is well-scoped for a server managing Antigravity skills, configuration, providers, and version info. Each tool has a defined role, and the count is appropriate without feeling bloated or sparse.
Completeness4/5
The skill lifecycle is well covered: list, read, search, install, delete, and GitHub-based install are all present, with config get/set and provider listing. Minor gaps exist such as no direct provider add/remove tool or config deletion, but these can be worked around via set_config.
Related MCP Connectors
Search, fetch, lint, and install Agent Skills (SKILL.md) from the SkillMD registry.
Search and discover Agent Skills from the skills.sh registry. Powered by HAPI MCP server.
Search & install 13,000+ AI agent skills from skills-hub.ai inside any MCP tool.
Governed AI agent skills — one library, distributed to devs and exposed to remote agents over MCP.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables managing local repositories, CLIs, MCP servers, and skills via MCP tools, with inventory tracking, guarded updates, and Docker-sandboxed trials.MIT
- AlicenseNot gradedqualityCmaintenanceEnables managing a canonical library of agent skills and MCP servers, syncing them across multiple development harnesses, and adding, importing, or configuring them through MCP tools.218 npmMIT
- AlicenseNot gradedqualityBmaintenanceEnables MCP-capable agents to search, inspect, lint, and safely install Agent Skills from the skillmd registry mid-conversation.20 npm143MIT
