Toolradar MCP Server: Give Your AI Agents Access to 8,600+ Software Tools

Ask Claude to recommend a project management tool and you will get a confident answer — often based on training data from 2024. The pricing will be wrong. The "free tier" it mentions may have been removed six months ago. And the tool that launched last week with 2,000 Product Hunt upvotes? It does not exist in the model's world.

This is the core problem with using LLMs for software recommendations. Their training data has a cutoff, typically 1-2 years behind reality. They compensate by sounding certain about things they are guessing.

We built the Toolradar MCP server to fix this.

What it does

The Toolradar MCP server gives AI agents — Claude Desktop, ChatGPT, Claude Code, Cursor, VS Code Copilot, Windsurf, Cline, and any MCP-compatible client — direct access to Toolradar's database of 8,600+ software tools. Instead of relying on stale training data, your agent queries live, structured data: verified pricing, G2 and Capterra ratings, editorial scores, feature lists, and AI-identified alternatives.

Six tools, one server:

Tool	What it does
`search_tools`	Search by keyword, category, pricing model. Returns ranked results with scores.
`get_tool`	Full details on any tool — pricing, pros/cons, ratings, description.
`compare_tools`	Side-by-side comparison of 2-4 tools with a "best overall" pick.
`get_alternatives`	Direct competitors for a given tool, identified by AI — not just category matches.
`get_pricing`	Detailed pricing breakdown with plan names, prices, and features.
`list_categories`	Browse all categories with tool counts.

Every response is structured JSON. No HTML, no markdown, no parsing required. Designed for agent consumption.

How it compares to web search

You might think: "Can't I just give Claude a web search tool?" You can. Here is why that produces worse results for software questions.

Web search returns marketing pages. Ask a search MCP server for "best CRM tools" and it fetches SEO-optimized listicles. The AI has to parse HTML, separate ads from content, and reconcile conflicting claims across ten articles. The result is a summary of other people's opinions, not structured data.

Toolradar returns structured facts. The same query via search_tools returns JSON with exact pricing tiers, verified ratings, editorial scores, and feature lists. No parsing. No opinion laundering. The AI can compare tools on specific dimensions because the data is already normalized.

Speed matters too. A web search workflow (search → fetch 3-5 pages → parse → synthesize) takes 10-20 seconds and burns 5,000+ tokens on HTML processing. A Toolradar compare_tools call returns a structured comparison in under 2 seconds.

Web search is the right tool for open-ended research ("what did TechCrunch say about Notion's Series C?"). Toolradar is the right tool for software decisions ("which CRM has the cheapest plan with API access and Slack integration?").

Sample queries and what comes back

Here is what actually happens when your agent calls Toolradar tools.

"Find me free CRM tools with API access"
→ Agent calls search_tools with q: "CRM", pricing: "free".
→ Returns 15+ tools ranked by editorial score, each with: name, slug, one-line description, pricing model, G2 rating, editorial score, and a link.

"Compare Linear vs Jira vs Asana for a 20-person startup"
→ Agent calls compare_tools with slugs: ["linear", "jira", "asana"].
→ Returns a side-by-side table: pricing per seat, free tier limits, G2 rating, editorial score, pros/cons, and a "best overall" pick with reasoning.

"How much does Vercel cost?"
→ Agent calls get_pricing with slug: "vercel".
→ Returns each plan (Hobby: Free, Pro: $20/user/month, Enterprise: custom) with feature breakdowns per tier. Not a guess — fetched from Vercel's actual pricing page within the last week.

"What are the best alternatives to Heroku?"
→ Agent calls get_alternatives with slug: "heroku".
→ Returns AI-identified direct competitors: Render, Railway, Fly.io, Vercel, Netlify — tools that solve the same deployment problem for the same audience. Not "AWS" or "Google Cloud" which are in the same broad category but serve different use cases.

Why this matters

Three things go wrong when LLMs recommend software without live data:

Stale pricing. SaaS companies change pricing constantly. A model trained in early 2024 will quote Notion's old pricing tiers. Toolradar verifies pricing pages weekly.

Missing tools. The AI development tools space moves fast. Tools like Cursor, Bolt, and Lovable went from zero to mainstream in months. If they launched after the training cutoff, the model does not know they exist. Toolradar adds new tools daily from Product Hunt, Hacker News, Techpresso, and community submissions.

Fake alternatives. Ask an LLM for "Figma alternatives" and you might get Canva — which is not a direct competitor. Toolradar uses AI analysis to identify actual competitors: tools that solve the same problem for the same users. Figma's alternatives are Sketch, Adobe XD, and Framer. You can see the full comparison at Figma vs Sketch or Figma vs Framer.

Setup in 2 minutes

1. Get a free API key

Sign in at toolradar.com/dashboard/api-keys and generate a key. No credit card required. The free tier includes 100 API calls per day.

2. Add to Claude Desktop

Open your Claude Desktop config file and add the Toolradar server:

{
  "mcpServers": {
    "toolradar": {
      "command": "npx",
      "args": ["-y", "toolradar-mcp"],
      "env": {
        "TOOLRADAR_API_KEY": "tr_live_your_key_here"
      }
    }
  }
}

Not sure where the config file is? See our full Claude Desktop MCP setup guide.

For Claude Code, it is even simpler:

claude mcp add toolradar -- npx -y toolradar-mcp

For Cursor, add the same config to your MCP settings. Any MCP-compatible client works — see our guides for Cursor and VS Code.

3. Start asking

Once connected, your agent can use Toolradar tools through natural conversation:

"Find me the best free CRM tools with API access"
"Compare Notion vs Clickup vs Asana"
"How much does Linear cost?"
"What are the top alternatives to Intercom?"

Claude will call the appropriate Toolradar tool, get structured data back, and give you an answer grounded in current information — not a guess from 2024.

What makes the data good

Most "tool databases" are scraped once and left to rot. Toolradar is different:

Pricing verified weekly. We fetch actual pricing pages and parse plan details. When Slack changes its per-user price, we know within days.
G2 and Capterra ratings. Aggregated review scores from real users, not LLM-generated summaries.
Editorial scores. Every tool is scored on functionality, ease of use, value, and support. Computed from multiple signals, not vibes.
Logos always current. Every tool has a validated, high-quality logo (48px minimum). Four layers of validation ensure no broken or pixelated images.
New tools daily. We track Product Hunt, Hacker News, Techpresso, and community submissions. Your agent knows about tools that launched this week.
8,600+ tools indexed. From major platforms (Slack, Figma, Notion) to niche developer tools and emerging startups. Browse all categories to see coverage.

REST API for non-MCP clients

The MCP server is a thin wrapper over our REST API. If you are building an agent with LangChain, CrewAI, or any HTTP-capable framework, you can call the API directly:

GET https://toolradar.com/api/v1/tools/search?q=project+management&pricing=free
GET https://toolradar.com/api/v1/tools/linear
GET https://toolradar.com/api/v1/tools/compare?slugs=notion,clickup,asana
GET https://toolradar.com/api/v1/tools/figma/alternatives
GET https://toolradar.com/api/v1/tools/figma/pricing

Full documentation at toolradar.com/docs. Understand the difference between MCP and REST APIs in our MCP vs API guide.

Use cases

Developer choosing a stack. "I need a database for a Next.js app — compare Supabase, PlanetScale, and Neon" → structured comparison with pricing, features, community scores. See a live example at Supabase vs PlanetScale.

Procurement team evaluating vendors. "Find project management tools under $15/user/month with SSO and Jira migration" → filtered search with specific plans and pricing. Browse the full project management category.

AI agent building a recommendation. An autonomous agent calling search_tools → get_tool → compare_tools in sequence to produce a structured recommendation with data behind every claim. This is the pattern most agent frameworks (LangChain, CrewAI, AutoGen) use for tool-augmented reasoning.

Content writer fact-checking. "Does Slack still have a free tier? What's the message limit?" → verified pricing data, not a guess from training data. Check Slack's pricing page on Toolradar.

Consultant preparing a shortlist. "My client needs an analytics tool. List options under $50/month that integrate with Shopify" → filtered results with pricing details, ready to paste into a recommendation deck.

Free to use

The free tier includes 100 API calls per day across all six tools. That is enough for most agent workflows — typical usage is 5-15 calls per conversation. No credit card, no commitment.

If you need higher limits, get in touch.

The full landing page is at toolradar.com/for-agents. Documentation at toolradar.com/docs. The npm package is toolradar-mcp.

Toolradar MCP Server: Give Your AI Agents Access to 8,400+ Software Tools