treg

treg is an open source, pay-per-call gateway that lets AI agents use 3,800+ SEO, enrichment, ads, social and scraping endpoints from 110 providers through one key and one MCP server, with no markup.

treg is a metered gateway that gives AI agents access to a large catalog of third-party data and action endpoints through a single credential. The catalog spans more than 3,800 endpoints from 110 providers, covering SEO and keyword research, backlinks, AI visibility tracking, social media data, people and company enrichment, buying signals, ad management, competitor creative and measurement. Instead of signing up with Semrush, Moz, Hunter, Apollo or Crunchbase separately, an agent holds one treg token and calls whatever it needs. The service is aimed at people running coding and automation agents such as Claude Code, Codex, OpenClaw, opencode and similar, who want those agents to act as an SEO analyst, social media manager, outbound sales rep or media buyer. Setup is a single instruction handed to the agent that points it at treg's llms.txt file. From there the agent can request a task rather than a named tool. For categories with many vendors, such as backlinks or email lookup, treg shows each provider's price alongside measured success rate, speed and last response time, so the agent can choose on evidence or pin a specific vendor. Billing is per call rather than per seat. Each call is priced before it runs, with examples on the homepage in fractions of a cent, and new accounts start with a $1.00 credit. The site states a 0% additional fee, with revenue coming from volume pricing agreed with vendors. Users who already pay for a provider can bring their own key, and those calls route through it unmetered. Credentials are injected server-side and never sit on the user's machine, and some account-bound tools such as Search Console, GA4 and ad platforms use OAuth. treg is open source under AGPL, and developers can list their own APIs or skills in the catalog. It sits between direct vendor subscriptions, which are costly for occasional agent use, and generic scraping or API-hub services, and the closest comparison is a model router like OpenRouter applied to tools and data instead of language models.