Skip to content
Conduit

Documentation

Conduit is a directory of tools for AI agents — MCP servers and agent skills. This page explains how connecting works and how to read the catalog programmatically.

Connect a tool

Open a tool and press Use. Choose where it runs, which agent you use and your operating system, then fill in any required settings. You get a config made for that agent. Values you type stay in your browser — they are only used to fill in the config.

For example, a local MCP server in Claude Code:

terminal
claude mcp add sql-explorer -e DATABASE_URL='<your DATABASE_URL>' -- npx -y @example-org/sql-explorer

Cloud and local

Local tools are started by your agent on your computer. They can reach your files and apps, and your data stays on your machine. Cloud tools run on a server and are reached through a URL.

Each listing keeps two facts apart: whether a tool can run in the cloud, and whether it is actually running on this platform. Cloud runtime on Conduit is not active yet, so today cloud connections only exist for tools their authors host themselves.

Statuses

Listed
The repository exists and its public README and license were read.
Unreviewed
No one has inspected the code. This is the default.
Reviewed
Someone inspected the code. The page says who and when.
Cloud-ready
Running on this platform and connectable by URL now.
Claimed
A maintainer proved they control the repository.

Costs

Using Conduit is free. Some tools call paid services — a search API, a cloud provider, a design tool. Those charges come from the service, under its own pricing, and the tool page says when that applies.

Public API

Read-only JSON, no key required. Results are paginated on the server.

GET /api/v1/tools
curl "https://conduits.dev/api/v1/tools?q=postgres&view=cloud&per_page=20"

# Parameters
q          free-text search (name, summary, tags, categories)
view       all | cloud | local
kind       all | mcp | skill
category   category slug, e.g. databases
license    SPDX id, or "none"
api_key    any | required | none
cloud      any | capable | ready | local-only
sort       relevance | recent | name
page       1-based page number
per_page   1–50 (default 12)
GET /api/v1/tools/{slug}
curl "https://conduits.dev/api/v1/tools/sql-explorer"

For AI agents

Agents can start at /llms.txt for a plain-text map of the site, or /llms-full.txt for every listing in one file. Pages are rendered on the server, include structured data (schema.org), and link to their JSON form with <link rel="alternate">.

↑↓ navigate openesc close