The failure this fixes is specific: a model quoting last year's price or inventing a comparison table. Here the pricing is verified weekly and the alternatives are identified rather than assumed from category membership. The free tier's 100 calls a day is enough for research, less so for anything automated.
An MCP server over Toolradar's software directory. It answers the questions models get wrong from training data — what a tool costs today, what it actually competes with, how it scores — against pricing checked weekly and G2 and Capterra ratings.
- search_tools — search the database by keyword, category or pricing model, with names, descriptions, pricing, scores and categories
- get_tool — one tool's full record: description, pricing, pros and cons, features, editorial score, review synthesis and alternatives
- compare_tools — two to four tools side by side, with computed insights for best overall, best value and most reviewed
- get_alternatives — up to ten competitors for a tool, sorted by editorial score
- get_pricing — tiers, pricing model, free-trial availability and an expert pricing verdict
- list_categories — every category with its tool count, which is how you find valid slugs for the search filter
A Toolradar API key in TOOLRADAR_API_KEY; a free key allows 100 calls a day, resetting at midnight UTC. Clients run npx -y toolradar-mcp over stdio. TOOLRADAR_API_URL points at a different endpoint if you need one.
One command plus a key — npx -y toolradar-mcp, then supply credentials