| name | find-agent-friendly-tools |
| description | Find agent-friendly tools with Can Agents Use when a user asks which software, SaaS, API, CLI, MCP server, or browser workflow an AI agent can actually operate safely. Use for general tool discovery, alternatives, shortlists, and evidence-backed recommendations. |
| allowed-tools | WebFetch |
Find Agent-Friendly Tools
Use this skill when the user asks for tools an AI agent can use, asks for alternatives in a software category, or wants a shortlist with evidence for API, CLI, MCP, browser support, docs, pricing, account setup, and sandbox readiness.
Fast Path
- Search Can Agents Use first:
curl -fsS 'https://canagentsuse.com/api/agent/search?q=<query>&page=1&limit=10'
- If the agent can use MCP, prefer the read-only MCP endpoint:
{
"mcpServers": {
"canagentsuse": {
"type": "http",
"url": "https://canagentsuse.com/api/mcp"
}
}
}
- Fetch a specific tool before recommending it:
curl -fsS 'https://canagentsuse.com/api/agent/tools/<slug>'
- For broad comparisons, fetch the complete catalog once and search locally:
curl -fsS 'https://canagentsuse.com/api/agent/catalog'
What To Return
- Tool name, slug, website, and docs link.
- Agent score and the strongest capability signals.
- Which interface the agent should use first: MCP, API, CLI, browser, or docs.
- Pricing, auth, account setup, sandbox, and caution notes.
- Evidence URLs for important claims.
Guardrails
- Do not scrape HTML when API, MCP, OpenAPI, or Markdown surfaces are enough.
- Treat scores as discovery signals, not security, legal, compliance, or purchasing approval.
- Require human review for payments, production data, account creation, infrastructure, compliance, or irreversible actions.
- Prefer official documentation and public evidence over marketing summaries.