Treg — The OpenRouter for Agent Tooling

Treg is a unified API gateway for agent tooling — a single entry point that gives you access to a growing catalog of third-party APIs without managing individual accounts, keys, or billing. The pitch that got us in: an OpenRouter for Agent Tooling.

We discovered treg through Bootoshi's announcement in the Vibe Coders University (VCU) Discord. Its founder AI Jason Z (JZ) had launched treg and was offering $30 in free credits to VCU members who filled a form and upvoted the Product Hunt launch. That promo credit powered an entire competitive intelligence system.

What We Built With $1

Before treg, our competitor table in Twenty was 312 records of raw noise — blog posts, docs pages, parked domains, and real competitors mixed together with no structure. After one session with treg's APIs:

• 250 clean competitor records — 50 noise records deleted, 35 reclassified, 19 duplicates removed, zero NOT_A_COMPETITOR left
• Everything classified — every record got a competitorType, competitiveLevel, and gtmPriority
• 250+ ranked keywords captured across 17 competitors (Anthropic, OpenRouter, Together AI, Fireworks AI, DeepInfra, and more)
• Keyword data live in Twenty — every competitor record has a rankedKeywords JSON field with weekly keyword snapshots

The data feeds directly into our comparison pages against competitors like OpenRouter, the AI gateway comparisons we publish against Portkey, and the side-by-side analysis we run against Together AI and Fireworks AI.

Total cost: $0.95 of the $1 promo credit. The $30 VCU grant is banked for the next pass.

How We Used Treg

Step 1: Competitor Discovery

We ran spyfu.google.domain.competitors on our top competitors through treg. For OpenRouter alone it returned 501 competing domains, each with a relevance score and shared keyword overlap — a ready-made market map in one API call.

Step 2: Content Verification

Many of the 312 raw records were blog posts, docs pages, and parked domains — not real competitors. We read the actual page content of each record with jina.ai/r (our Tier-1 web ingestion reader), then reclassified or deleted them: 50 noise records removed, 35 reclassified as real products.

Step 3: Keyword Capture

We ran serpstat.google.domain.ranked_keywords on our top 20 competitors. Each call returned the competitor's top-ranking keywords with position, search volume, keyword difficulty, and traffic — stored in Twenty's rankedKeywords field as a weekly snapshot.

Step 4: Classification

With clean data in hand, we classified every competitor into tiers. The TIER_1 roster that drives our VS-page strategy: Anthropic, SiliconFlow, Fireworks AI, DeepInfra, Together AI, Portkey, OpenRouter. Every record carries a competitorType, competitiveLevel, and gtmPriority — all stored directly in Twenty.

The Architecture

Treg fits into our broader GTM engineering stack as the data layer — the source of external market intelligence that feeds into our Twenty CRM, Buzz channels, and content strategy agents.

The data flows through Buzz CLI for structured event routing, Parallel CLI for multi-source competitive research, and DeepSeek V4 Flash via the InferenceSaver API for analysis at affordable inference pricing.

Treg (API gateway)  →  Buzz CLI (routing)  →  Twenty CRM (storage)  →  Agents (analysis)

Why treg instead of individual API accounts

• One billing account — no $5 minimums on five different platforms
• One authentication — a single treg key for every endpoint
• Credit-based pricing — $0.0002 per SpyFu competitor result, $0.0005 per Serpstat keyword. We spent $0.95 on 250+ keyword captures and discovery
• Structured JSON back — every endpoint returns data a pipeline can consume directly, no scraping

What's Next

With the $30 VCU grant banked, the next steps are:

1. Monthly keyword refreshes — re-capture ranked keywords for the competitor roster
2. Competitor discovery expansion — run spyfu on more TIER_1 and TIER_2 competitors
3. Content gap analysis — compare our keyword rankings against competitors' to find high-volume targets for VS pages
4. Automated market map — a weekly workflow that discovers new competitors, reads their content, and updates the table