The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Company Enrichment listing page.
One install gives your agent the superagnt_ MCP server — every data source (X, TikTok, YouTube, Instagram, Reddit, LinkedIn, web), people + company enrichment, a real workspace Postgres, schedules, webhooks and shareable canvases — plus the skills that teach the workflows.
Auth is OAuth on the first tool call; the consent screen doubles as signup. No tokens, no config editing, no secrets in this repo.
Skill plugins are packaged workflows: the skill plus the tools it needs. Each pulls the base plugin in as a dependency, so one line installs the whole stack:
Every skill has a page describing exactly what it does before you install anything: https://superagnt.com/skills
plugins/superagnt/ — the MCP server config plus eight base skills
(data-sources, lead-generation, workspace-db, task-agents, automations,
data-pipelines, canvas, superagnt-setup) and the /superagnt:connect
command.plugins/<skill>/ — one skill per packaged workflow. Read any skill before
installing; they are plain markdown.Two rules every skill follows: your agent can propose a paid module (each has a free trial) but money only ever moves on a confirmation page you open yourself, and nothing sends outbound (email, posts, DMs) — drafts are the deliverable.
Not on Claude Code? The same setup is one paste in any MCP client:
Generated from the superagnt monorepo (tools/claude-plugins/) — issues and
PRs here are read, but content changes land upstream and sync out. MIT
licensed.