skillfed
for your agent

Connect your agent.

Every way in, on one page: the MCP server, the CLI finder, the Python CLI, chat search, and the machine-readable API. Wishes in — curated, deduplicated matches out.

Quick start

pick your client
# one line — registers the skillfed MCP server claude mcp add skillfed -- npx -y skillfed-mcp
# .cursor/mcp.json, or VS Code settings → mcp.servers { "mcpServers": { "skillfed": { "command": "npx", "args": ["-y", "skillfed-mcp"] } } }
# stdio transport · Node ≥ 18 · zero config, defaults to skillfed.io command: npx args: -y skillfed-mcp
# Python CLI + finder skill pip install skillfed skillfed --help
# no install, no account — any AI chat # your chat writes an abstract wish as a plain-English URL, # you approve it, it reads five security-swept results: https://skillfed.io/api/q/<what-you-need>

No agent, just a chat window? Full walkthrough: using SkillFed from a chat.

Wishes stay abstract by design — a wish describes the capability you need, so your code, files, and plan never leave the machine. Node ≥ 18 for the npx paths.

What your agent gets

skillfed-mcp 0.2.3

find_skills

the discovery path

Lexical-recall search over a wish-list — abstract wishes in, deduplicated skill matches out, every match a live page.

01skills

find_packages

the capability index

Capability search over the PyPI package index — license, maintenance, and install-friction facts locked on every card.

02packages

find_research

the evidence

Topic search over 191 claim-first research notes on the agent-skills literature.

03research

get_skill_bundle

the follow-through

Pulls a matched skill's full text for your agent to read in context — writing it to disk is a separate, later, approved step.

+bundle