The dependency picker for coding agents

Before your agent adds a library, it asks whichlib. whichlib finds the candidates on GitHub and ranks them by a transparent score: real momentum, maintenance, adoption and license.

Install

Nothing to install. Claude Code:

claude mcp add --transport http whichlib https://whichlib.com/mcp

Cursor, VS Code and other clients: see the docs. Free for 50 calls a day; add your own GitHub token for unlimited daily use.

Connects to the hosted server and adds a skill that makes Claude check a library before adding it:

/plugin marketplace add josifb/whichlib
/plugin install whichlib@whichlib

Nothing runs on your machine. Free for 50 calls a day; add your GitHub token in the plugin settings for unlimited daily use.

Runs on your machine, no limits beyond GitHub's own (Node 22 or newer):

claude mcp add whichlib -- npx -y whichlib

Cursor, Windsurf and others: { "mcpServers": { "whichlib": { "command": "npx", "args": ["-y", "whichlib"] } } }

“Python http client”

A real answer, saved from the hosted API on .

  1. psf/requestsStrong 81

    Gaining steadily, pushed 16 days ago, Apache-2.0.

    Fit 7954,460 stars

  2. encode/httpxStrong 79

    Gaining steadily, pushed 12 days ago, BSD-3-Clause.

    Fit 7915,603 stars

  3. urllib3/urllib3Solid 62

    Slow growth, pushed 6 days ago, MIT.

    Fit 594,071 stars

Fit is score × relevance to the question. Every part of the score comes back with the answer.

How the score works

Momentum
Stars gained this week from daily star counts (the top 1,000 repositories per language plus new ones). Other repositories fall back to stars per week, adjusted by their npm or PyPI download trend.
Maintenance
How recently the repository was pushed to, with a floor for widely used libraries that are simply stable.
Adoption
Stars, forks and weekly npm or PyPI downloads, counted only when the registry links back to the repository.
License
Permissive licenses score highest; no license scores zero.

Nobody can buy a better rank: the data is public GitHub, npm and PyPI data, and the formula is open source. The details are in the docs.

Teams

Shared limits, an allow-list for your organisation and usage reports are coming. Join the waitlist (a thumbs-up on the issue is enough).