treg

treg is the tool catalog: about 2,600 endpoints across about 40 providers (web search, X and TikTok data, SEO, people and company enrichment, ad libraries...), each called with the key injected server-side, so nobody holds the provider's key. log uses it for every lookup that has to come from an external source, never from memory or a guess (rules 53, 166).

What log can do with it

ToolCommandWhat it gives
tregSearchlog treg search "<what you want done>" [--limit <n>]The endpoints that do the job, each with its id, what it does and its price
tregGetlog treg get <endpointId>One endpoint in full: its price, every param and a call that works
tregCalllog treg call <endpointId> --params '<json object>' [--path <p>] [--method <m>] [--data '<body>'] [--await]Its answer, as the provider sent it
tregCallLivethe same with --live --approvedBy mattThe same, for a priced endpoint matt asked to spend on

What it costs

Every call reads the endpoint's price with get first:

  • A free endpoint runs. tinyfish.web.search is free, for example.
  • A priced endpoint is a dry run. It prints the price and the exact treg call it would run, and calls and spends nothing.
  • It runs only with --live and --approvedBy, when matt asked for that spend (rule 41). Over MCP that's tregCallLive with confirmLive: true.

The team's treg balance is in log balance (rule 164).

How it runs

log runs the treg CLI on this machine (treg.exe, found on the PATH; TREG_BIN moves it), as a child process with its arguments passed one by one, never through a shell. Each search, get and call is a timed span in the call's timings (treg.search, treg.get, treg.call, rule 165), so log timing shows how long treg took.

Example

log treg search "web search"
log treg get tinyfish.web.search
log treg call tinyfish.web.search --params '{"query":"kick streamer managers","domain_type":"web"}'

Setting it up

Install the treg CLI and sign in (treg --help). There is no key in .env.local: treg holds the credentials server-side.

Spec: treg.