Skip to main content
find_tools is the entry point for every AgentKey workflow. It performs semantic search across all 1,800+ tools using a combination of embedding similarity, platform-alias boosting (so mentions of “X”, “Twitter”, or ”𝕏” all resolve correctly), and intent-keyword boosting (so action verbs like “search”, “scrape”, and “trending” carry extra weight). The result is a ranked list of Provider/Operation names sorted by relevance to what the user actually wants to do.
​
q
string
required
The user’s original, unmodified phrasing — not a single extracted keyword. The full sentence gives the router both the action verb and any platform context it needs for accurate ranking. Works in English, Chinese, or mixed language. Examples:
  • "search the latest OpenAI news"
  • "find trending posts on X about GPT-5"
  • "scrape the article at https://example.com"
  • "获取比特币实时价格" (fetch the real-time Bitcoin price)

​
Returns

A ranked list of matching tools. Each item in the list contains:
​
name
string
The canonical Provider/Operation identifier for the tool, for example Twitter/searchTweets or Firecrawl/scrape. Pass this value directly to describe_tool or execute_tool.
​
summary
string
A short, plain-language description of what the tool does.
​
cost
object
Per-call pricing for the tool.
​
cost.credits_per_call
number
The number of credits deducted from your balance each time the tool is executed.

​
Example

find_tools("what people are saying about Claude on X today")

[
  { "name": "Twitter/searchTweets", "summary": "Search recent public tweets", "cost": { "credits_per_call": 0.5 } }
]
Do not pre-extract a single keyword from the user’s message before passing it to find_tools. Passing the full intent — including the action verb, any platform name, and any modifiers like “latest” or “trending” — gives the router everything it needs to surface the most relevant tools. A query like "X" will return far noisier results than "find trending posts on X about AI".
If you already know the category you want to browse, use list_tools(prefix) instead. It walks the category tree and returns all tools under a given prefix. Available top-level prefixes include search, scrape, social, crypto, finance, business, and ecommerce.

​
Next step

Take the name of the tool you want and pass it to describe_tool to get its full parameter schema, per-call cost, and a ready-to-run template.

describe_tool

Get a tool’s parameters, cost, and execute_as template before you run it.