mh install, wherever you already work.Install mh once; from then on mh install works in any harness. 11 editors and counting.
curl -fsSL https://metahub.ai/install.sh | shReadable script · installs mh + MCP server via your package manager · commit-pinned installs · opt-out telemetry, never your prompts
...and any other tool that speaks SKILL.md or MCP.
Every editor speaks a slightly different language. metahub is the layer in between: one catalog of skills, MCP servers, agents, and plugins that install once and run everywhere, with free observability for the people who build them.
Browse a curated catalog, or search by name, tag, or author; the best, verified picks rise to the top. Every listing shows reviews, freshness, and the exact commit it's pinned to, so you install with confidence.
Every listing is pinned to a specific commit. What you install is exactly what was tested. Updates land when you choose. Works across Claude Code, Cursor, VS Code, Zed, and 8 more editors.
The install command above also connects metahub to your editor’s assistant: Claude Code, Cursor, Claude Desktop, Continue, and any editor that speaks MCP. Nothing else to set up. Describe what you need, and your assistant finds it, installs it, and reports back.
“find me a skill for parsing PDFs”
Searches the catalog in plain language and ranks results by real installs and ratings.
metahub_search“install that”
Installs what you just found straight into your editor. No terminal step; ask for the command instead if you prefer to run it yourself.
metahub_install“what do I have installed?”
Answers from your machine, not the cloud: it reads the local install ledger at ~/.metahub/installs.json.
metahub_list_installed“sign me in”
Hands you a GitHub link to approve in your browser. Signing in unlocks reviews and publisher stats.
metahub_signin_begin“leave a 5-star review on the pdf skill”
Posts a verified review to the tool's public listing. Requires sign-in.
metahub_submit_review“how is my pdf-parser doing?”
Live install counts, latency, and errors for tools you published. Requires sign-in.
metahub_my_statsPublisherSo we run it. Every artifact is executed in a sandbox against real prompts and graded on the transcript (correctness, instruction-following, safety, latency), then probed with prompt-injection and destructive requests. The rubric and the judge prompts are open source as Assay.
Run in a sandbox against real prompts, graded on the transcript, then probed with prompt-injection, out-of-scope and destructive requests. Every number is pinned to the commit above.
Connect a GitHub repo, run the checks, publish. Real reviews from verified installs. An open SDK that wraps your artifact in one line for free telemetry.
Sign in with GitHub. Pick a repo with a skill, MCP, agent, or plugin.
Structure, security, and quality scans.
Your artifact is in the registry. Observability starts immediately.
Publish a GitHub repo, get a live dashboard. Invocations, p95 latency, top models, handoffs, error rate. One-line SDK wrap, no config, no cost.
Reviews come only from developers with a verified install, so the score reflects real usage, not drive-by stars.
The highest-rated picks across skills, MCP servers, agents, and plugins.
Verification Before Completion
@obra
Evidence before assertions, always
Open Webui
@open-webui
Open WebUI
Superpowers
@obra
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Claude Agent Sdk Python
@anthropics
Python SDK for Claude Code
Writing Plans
@obra
Turn specs into phased implementation plans
openclaw
@openclaw
Multi-channel AI gateway with extensible messaging integrations
Both take 90 seconds or less. Both are free.