If youâre a data engineer figuring out how your pipelines, models, and metadata actually serve AI systems in production, OpenXData is running a free virtual conference built around exactly that.
April 29, 9amâ3pm PT.
Technical sessions and workshops that go beyond design slides to what holds up in production. From engineers at Uber, Meta, Booking, Snowflake, Anthropic and many others.
The kind of problems we talk about here every week, discussed by the teams solving them at scale.
If your skills live as local files, only YOUR client benefits.
Publishing them through MCP makes that knowledge available to every connected client, on any platform.
For example, this is how Claude looks with Substack Author MCP:
This is how Claude allows you to load skills that come from the MCP server. The manifest files are exposed config files that are coming from FastMCP but donât need any interaction to work.
If you want more context around SKILLs and the MCP server I built for Substack, feel free to take a look below.
The server runs on FastMCP. Python functions become MCP tools with decorators, type hints become JSON Schema automatically, and one line starts the HTTP transport.
What matters for this article: substack-author-mcp currently exposes 8 tools and 5 skills. Any MCP client connects and gets both.
Hereâs the repo structure:
server.py registers the tools AND points SkillsDirectoryProvider at .claude/skills/. Both get exposed through the same MCP endpoint.
And this is how it looks when you use them:
You can see how Claude loads the skills and uses the MCP tools to get the data.
This is the whole point. Two lines on the server side:
from fastmcp.server.providers.skills import SkillsDirectoryProvider
if settings.skills_path.exists():
mcp.add_provider(SkillsDirectoryProvider(roots=settings.skills_path))
SkillsDirectoryProvider scans the directory (in this case .claude/skills/ is configured as settings.skills_path), finds every SKILL.md file, and registers each one as an MCP resource with a skill:// URI. Any MCP client can then:
Call list_resources() to discover available skills
Call read_resource("skill://analyze-notes/SKILL.md") to read the full instructions
Read the manifest (skill://analyze-notes/_manifest) to discover supporting files
The same SKILL.md files I use locally in Claude Code are now exposed as accesible resources.
On the client side, FastMCP provides utilities to consume those skills programmatically:
from fastmcp import Client
from fastmcp.utilities.skills import list_skills
async with Client("http://skills-server/mcp") as client:
skills = await list_skills(client)
for skill in skills:
print(f"{skill.name}: {skill.description}")
# Download one skill
path = await download_skill(client, "pdf-processing", Path.home() / ".claude" / "skills")
# Or download all skills
paths = await sync_skills(client, Path.home() / ".claude" / "skills")
You point a client at your MCP server and it gets the tools AND the instructions for how to use them, thereâs no need to share prompts, thatâs so 2024 đ€Ł
FastMCP ships with vendor-specific providers that already know where each platform stores skills on their filesystems:
Skills arenât locked to one platform. FastMCP treats them as MCP resources regardless of origin. You could serve Claude skills to Cursor clients, or Gemini skills to a custom agent with the same format: a directory with a SKILL.md file.
This is the real play. Skills were born inside Claude Code, but the format is just markdown in a folder. FastMCP makes that format a protocol-standard resource that any MCP client can consume.
If you enjoyed the content, hit the like â€ïž button, share, comment, repost, and all those nice things people do when like stuff these days. Glad to know you made it to this part!
Hi, I am Alejandro Aboy. I am currently working as a Data Engineer. I started in digital marketing at 19. I gained experience in website tracking, advertising, and analytics. I also founded my agency. In 2021, I found my passion for data engineering. So, I shifted my career focus, despite lacking a CS degree. Iâm now pursuing this path, leveraging my diverse experience and willingness to learn.
