Treg review (2026): one gateway for thousands of APIs, unchecked data

What is Treg?

Treg is a catalog of APIs for AI agents: 4,000+ endpoints from 113 providers, from SEO and social data to people and company enrichment, behind a single token. Superdesign, the team behind it, publishes the code on GitHub. You never sign up with the providers. Treg calls them with its own keys, or with keys you add, and bills a prepaid balance per call.

Most of Treg is a relay: a call to a provider endpoint returns that provider's response unchanged. For a few common jobs, mostly finding an email or a phone number, Treg has routed endpoints. Ask treg.people.email.find or treg.people.phone.find for a person, and Treg tries the providers that can answer one at a time until one returns something. It ranks them on every call by expected cost per hit, your own keys first, using hit rates it measures across all its users. The response names the provider that answered, every provider tried before it, and what each one cost.

How it did in our test

We ran Treg's routed lookups live on LinkedIn profiles from our provider benchmark and compared them with pipe0's default waterfalls on the same people.

On coverage, Treg holds its own. It returned a mobile number for 94 of 97 people and a work email for 103 of 120, about as many as our waterfalls. Its first provider, Blitz, is fast and accurate: it answered half the email lookups in under a second, and those answers matched other providers 96% of the time.

The problem is what comes back when the first providers miss. Treg's waterfall takes the first answer from whichever provider is next in price order, and some of them return data that is plainly wrong.

Of Treg's 94 mobile numbers, 12 could not be dialled as mobiles: six were invalid or too short, three could not be parsed, two were landlines and one was a business line. Six of the twelve came from Aviato, which often returns US numbers without the +1, and Treg passes them on unchanged.

On email, 8 of 103 answers were personal mailboxes such as Gmail, Yahoo and Hotmail. LeadMagic's personal email finder sits inside Treg's work email route, and Tomba returned the rest.

After removing those, Treg's usable coverage was 85% on mobile numbers and 79% on work emails, against 94% and 83% for pipe0's default waterfalls.

Do they return the same data?

Often not. For the 94 people where both Treg and pipe0's waterfall returned a work email, they returned different addresses for 35 of them. For mobile numbers, 25 of 94 were different. When two waterfalls disagree on the same person, at most one of them is right.

How often a provider agrees with premium sources is the best signal of accuracy we have. Treg's mobile numbers matched Amplemarket's 72% of the time, a typical rate: Wiza, Prospeo and LeadMagic land at 72 to 73%. With Surfe it was 69%, lower than most providers we test, which match Surfe 77% of the time or more. And Treg agrees with low-cost providers about as often as with premium ones: nothing in its routing pulls its answers toward the sources we trust most.

Speed

Speed depends on how far down the list a lookup goes. When the first provider answers, Treg returns in about a second. Each miss adds the next provider's time, and some of them are asynchronous, so a lookup that falls through five or more providers averaged 47 seconds in our email test, and 7 of 120 ran past our 2-minute timeout. Mobile lookups were faster, at 7.1 seconds on average.

What it costs

There is no subscription. New accounts get $1 of credit, and after that you top up a prepaid balance from $10. Each answer costs exactly what the provider that found it charges, from $0.0005 for Blitz to $0.75 for Lusha on phones, and Treg adds no markup. Misses are free on all phone providers and all but one email provider, and a cost cap per call stops a lookup from reaching the expensive end of the list.

For mobile numbers, that came to $0.11 per usable number in our test, below what pipe0 charges for its default waterfall. If you already pay for a provider, you can add your key and those calls run unmetered.

Treg for agents

The gateway itself is well built. An agent sets itself up from one line, set up treg - https://treg.to/llms.txt, and gets an MCP server, a CLI and a guide written for it. Every call can carry a cost cap and an idempotency key, prices are shown before you spend, and an out-of-balance call is refused with a clear message instead of failing halfway.

What sits behind it is less agent-friendly. Treg forwards each provider's response as it is: nothing is filtered, shortened or made uniform for an agent's context window, so the agent reads every vendor's full payload in every vendor's format. The real gain over separate MCP servers, or over teaching an agent each provider's API, is the single gateway: one token, one bill and one way to call.

Documentation is thin for the same reason. Each endpoint has a catalog entry with its input fields, an example call and a link to the provider's own docs, but no Treg page that explains how to use it, and no description of what comes back, because what comes back is the provider's raw response.

The bundling has a cost too. Thousands of endpoints sit behind one token, and apart from the few routed lookups, each returns its provider's own JSON in its own shape. An agent has to choose among all of them, and a team that wants a strongly typed API client has to model every response it uses itself. For a large list, the agent also stores, joins and re-runs every result on its own.

How Treg differs from pipe0

Treg and pipe0 overlap on waterfalls, but they are built for different jobs. Treg is plumbing around a call: one token, billing, key management, a cache of past answers shared across its users, and the provider's raw response. Nothing in it stores the people or companies you look up, so your agent keeps the results, joins them to your list and decides when to look again.

pipe0 is a stateless enrichment API too, but it wraps retrieval in a data model. Every lookup returns typed fields in the same shape whichever provider answered, and an agent can chain lookups, checks and transformations into multi-step automations in a single request. Default waterfalls are ordered for answer quality rather than price. Sheets are an optional cloud environment on top, for lists that need to persist or run on a schedule. Treg's catalog is far broader, with long-tail APIs from TikTok data to text to speech. pipe0's is narrower and goes deeper on sales data.

Who should use Treg

Use Treg if you are building an agent that needs many data sources behind one token, and you check what comes back. Its catalog, pricing and tooling are hard to beat for that.

Look elsewhere if the data goes straight into outreach. A waterfall ranked by cost needs a check after it, for valid numbers and real work addresses. pipe0 runs its waterfalls in a quality-first order, and lets an agent chain the checks that catch these errors into the same request.