Get started

Add to Your Product

You already have an AI agent in your product — a customer service bot, an internal assistant, a SaaS copilot, or something else entirely. ToolRouter gives that agent on-demand access to a massive and growing catalog of capabilities through a single integration. Image generation, video production, voice synthesis, web search, SEO analysis, company data, financial data, security scanning, document processing, and much more — 143+ tools and 745 skills, all available through one connection.

No per-tool setup. No managing API keys for dozens of providers. No maintenance when new tools are added.

Why add ToolRouter?

Your agent is smart, but it is limited to what you have built for it. When a user asks your agent to generate an image, research a competitor, check a website's performance, look up a company, or transcribe audio — it cannot do those things without custom integrations for each one. Building and maintaining those integrations is expensive, slow, and never-ending. And every new provider means another API key to manage, another billing relationship, another set of docs to read.

ToolRouter replaces all of that with one connection:

  • One integration, massive capabilities. Connect once and your agent can discover and call any of the 143+ tools in the catalog. You do not build, host, or maintain any of them. No API keys to manage — ToolRouter handles all the provider connections.
  • The catalog grows automatically. New tools are added regularly. Your agent gets access to them the moment they ship — no code changes, no redeployments, no configuration on your side.
  • Consistent interface. Every tool returns the same response shape. Your agent learns one format and it works for everything from web search to video generation.
  • Usage-based billing. Pay only for what your agent uses. No monthly subscriptions per tool, no minimum commitments.
  • Built-in reliability. Tools are backed by multiple providers with automatic failover. If one provider is down, the call routes to the next. All communication is over HTTPS with authenticated bearer tokens.

What tools does your agent get access to?

The catalog spans 12+ categories. Here is a snapshot of what is available today:

  • Media generation: Image generation, video production, video editing, video upscaling, voice synthesis, voice transformation, sound effects, music generation, audio dubbing, background removal, image upscaling, diagram generation, QR codes
  • Research and search: Web search, deep research, academic papers, book search, Wikipedia, tech news, web archiving
  • SEO and marketing: Site auditing, page speed analysis, keyword research, competitor analysis, site crawling, App Store ASO, Play Store ASO, ad library search
  • Company and people data: Company lookup, lead finding, people search, social profiles, name enrichment, email validation, brand extraction
  • Finance: Stock market data, SEC filings, crypto prices, currency exchange, economic calendar, financial calculators, tax reference, world economy data
  • Social media: Content creation, social search, comment extraction, trending content, social commerce, UGC content
  • Travel and events: Flight search, flight status, travel booking, hotels, local events, train tracking, public holidays
  • Security: HTTP security scanning, vulnerability scanning, penetration testing, vulnerability database
  • Documents: PDF processing, Excel tools, Word documents, PowerPoint tools
  • Geospatial: Geocoding, IP geolocation, EV charger locations, country data, timezone conversion
  • Science and reference: Chemistry lookup, ocean data, space data, night sky, earthquake monitoring, drug info, nutrition data
  • Lifestyle: Recipe finder, workout planner, grocery prices, sports scores, movie and TV search, trivia

The full catalog is browsable at toolrouter.com/tools. New tools are added every week.

What does this look like in practice?

Here are examples of what becomes possible once your agent is connected:

  • A customer service agent can search the web for answers, look up a customer's company information, check if their website has technical issues, generate screenshots of problem pages, or transcribe a support call — all within the conversation, without switching tools.
  • An internal assistant can generate images for a pitch deck, run SEO audits on your marketing site, pull stock market data for a report, translate documents, or research competitors — on demand, for the whole team.
  • A SaaS product with an AI copilot can offer capabilities like web scraping, security scans, document processing, lead enrichment, or video generation as built-in features — without building any of them.
  • A workflow automation can chain tools together as pipeline steps — scrape a URL, enrich the data with company info, generate a chart, and produce a PDF report.

Your agent decides *when* to use a tool. ToolRouter handles the *how* — finding the right tool, routing the API call, formatting the response, and tracking usage.

How do I integrate?

There are three ways to connect, depending on your setup. The simplest option requires no code at all. If your platform does not support MCP, the REST API works from any language.

Option 1: MCP (simplest — no code required)

MCP (Model Context Protocol) is the standard way AI agents discover and call external tools. If your agent platform supports MCP, point it at ToolRouter's hosted endpoint:

Server URL: https://api.toolrouter.com/mcp

Your account is created automatically on first use — no API key setup required. Your agent immediately sees the full catalog and can start calling tools.

This works with any MCP-compatible platform, including Claude, ChatGPT, Microsoft Copilot Studio, and custom agent frameworks.

Option 2: REST API (works with anything)

For full control, call the REST API directly from your backend. This works from any language, any framework, any platform.

1. Get an API key (one-time, automatic):

bash
curl -X POST https://api.toolrouter.com/v1/auth/provision

This returns a tr_live_* API key instantly. No signup form — save the key and use it for all subsequent requests.

2. Discover tools your agent can use:

bash
curl https://api.toolrouter.com/v1/tools

3. Call a tool from your agent's backend:

bash
curl -X POST https://api.toolrouter.com/v1/tools/call \
  -H "Authorization: Bearer tr_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "tool": "web-search",
    "skill": "search",
    "input": { "query": "best CRM for small business" }
  }'

Every tool returns the same JSON shape:

json
{
  "status": "success",
  "output": { "results": [...] },
  "usage": { "cost": 0.002, "credits_remaining": 9.99 },
  "meta": { "tool": "web-search", "skill": "search", "latency_ms": 380 }
}

Your agent parses one response format regardless of which tool it called.

Option 3: npm package (Node.js environments)

If your agent framework runs on Node.js, the toolrouter-mcp package acts as a local MCP server that connects to ToolRouter's hosted API. Your framework talks to it over stdio, and it handles auth, discovery, and execution.

Add it to your agent's MCP configuration:

json
{
  "mcpServers": {
    "toolrouter": {
      "command": "npx",
      "args": ["-y", "toolrouter-mcp"]
    }
  }
}

This is ideal for custom agent frameworks, LangChain or LangGraph setups, or any Node.js backend where your agent framework expects MCP-style tool discovery.

How does my agent discover what tools are available?

Your agent does not need a hardcoded list of tools. It discovers them at runtime:

  • Via MCP: The agent calls tools/list and gets back every tool with its name, description, and input schema. It picks the right tool for the task automatically.
  • Via REST: Call GET /v1/tools for the full catalog, or GET /v1/tools/search?q=seo to search by keyword. Each tool response includes skills, input schemas, and examples the agent can use to construct its call.
  • Via the discover meta-tool: Agents connected via MCP can call discover with a keyword (e.g., discover("web scraping")) and get back matching tools instantly.

This means your agent can respond to requests it has never seen before. A user asks for something new, the agent searches the catalog, finds a matching tool, and calls it — without you writing a single line of tool-specific code.

How do I control costs?

  • Dry runs: Add "dry_run": true to any call to preview the cost before executing. No credits consumed.
  • Budget limits: Set a monthly budget cap via the API so your agent cannot overspend.
  • Usage tracking: Call GET /v1/usage to see cost breakdowns by tool, or GET /v1/usage/history for individual call records with latency and cost per call.

What do I need to get started?

No signup forms. No configuration dashboards. No per-tool onboarding.

  1. Pick your integration method — MCP if your platform supports it, REST API if it does not.
  2. Connect your agent — point it at https://api.toolrouter.com/mcp or provision an API key with one HTTP call.
  3. Let the agent discover tools at runtime — it searches the catalog and picks the right tool for each request automatically.

Your agent connects, discovers, and calls 143+ tools through one interface. The catalog grows every week, and your agent gets access to every new tool without any changes on your side.

Where to go next

  • Quickstart for the shortest path to a working call.
  • API Reference for the complete REST endpoint documentation.
  • Integration for auth details, rate limits, async jobs, and asset delivery.
  • Billing for credits, pricing, and budget controls.