Your coding agent can pull live Google results, keyword volume, difficulty and backlink data through one key, and you pay for the calls it makes instead of a seat you use twice a month.
Live results — organic, news, maps — plus keyword data, backlinks, Search Console and GA4, all answering the same token. The organic results endpoint has served 2,075 calls at a 100% success rate.
| The old way | With treg | |
|---|---|---|
| What you pay for | Semrush $139/mo + SerpApi $75/mo + Moz $99/mo, running whether the agent does or not | One prepaid balance. A keyword lookup is a fraction of a cent |
| Keys | One account, one login and one API key per provider, spread across machines and .env files | One treg token. Every tool in the catalog answers to it |
| Picking a provider | You guess, or you use the one you already pay for | catalog get lists every provider for that job with price, measured success rate, median speed and when it last answered |
| Commitment | Annual plans and seat minimums to answer one research question | No subscription. Stop calling and you stop paying |
| The workflow | Export a CSV from the keyword tool, another from the rank tracker, join them by hand | One agent run: keywords, difficulty, live results and competitors in a single pass |
Paste this into the same agent. It installs the CLI, signs you in and registers the tools. One line, once — already set up? Skip to step 2.
set up treg — https://treg.to/llms.txt
Research the market for [product]. Find 50 relevant keywords with real search volume, difficulty and search intent. Identify the 10 best opportunities and show which competitors currently rank.
The agent searches by the job, not the vendor. catalog_search "keyword search volume" returns the
endpoints that do it, each with a price and whether treg.to can serve it without a key of yours.
It compares, then picks. Five providers answer this one capability, and the spread is not small:
| Provider | Cost | Success rate | Median | Notes |
|---|---|---|---|---|
| $0.0005 per keyword returned | 100% (77 calls) | 1.2 s | up to 1,000 keywords per call | |
| $0.00179 per keyword returned | 92.3% (20 calls) | 1.7 s | up to 5,000 per call | |
| $0.09 flat per call | 98.8% (165 calls) | 3.9 s | same price for 1 keyword or 1,000 — batch it | |
| free with a connected account | not yet measured | — | your own OAuth | |
| 10 API units per row | not yet measured | — | own key only |
Those numbers are what treg.to has actually observed across real calls, sample size included — a 100% over 8 calls is weaker evidence than a 99% over 121, and the catalog shows you both so you can tell.
It calls, then joins the results. Volume and difficulty for the list, live Google results for the shortlist, and authority metrics for whoever ranks (SE Ranking $0.002685, Moz $0.006667). One report, one bill.
The results endpoints are the most-used tools in the whole catalog, so their records are the longest:
| Result type | Provider | Cost | Success rate | Median |
|---|---|---|---|---|
| Organic | $0.002 | 100% (2,075 calls) | 5.5 s | |
| $0.00188 | 100% (343 calls) | 1.6 s | ||
| $0.0005 | 100% (77 calls) | 1.2 s | ||
| $0.015 | — | — | ||
| Maps / local | $0.002 | 100% (2,646 calls) | 1.9 s | |
| News | $0.002 | 100% (690 calls) | 9.0 s | |
| Trends | — | 98% (462 calls) | 6.7 s |
50 keywords · US · Google · pulled [date] KEYWORD VOL/MO KD INTENT TOP 3 RANKING <keyword> 8,100 61 commercial <domain> · <domain> · <domain> <keyword> 2,900 48 commercial <domain> · <domain> · <domain> <keyword> 1,300 34 informational <domain> · <domain> · <domain> ... THE 10 OPPORTUNITIES (volume weighted by difficulty and intent) 1. <keyword> 1,900/mo · KD 22 · commercial · weakest ranker DA 31 2. <keyword> 880/mo · KD 18 · commercial · nothing targets this exactly ... PROVIDERS USED serpstat (volume + difficulty) · dataforseo (results) COST $[from your run]
Structure is illustrative. The values are the providers' own, relayed unchanged.
Keyword research is a job you do once. Watching results is a job you do every week, and it is what the agents on treg.to overwhelmingly spend their calls on — the organic, maps and news result endpoints together serve more traffic than every other SEO tool in the catalog combined.
Every Monday, check where we and our top 3 competitors rank for these 20 keywords — organic, news and maps. Report only what moved since last week, and for anything that dropped, show which page overtook us.
Twenty keywords across three result types is 60 calls, about $0.12 at DataForSEO's $0.002. Run weekly for a year and it costs roughly $6 — against a rank tracker at $65/mo. And the week you skip it, it costs nothing, which is the part a subscription can never do.
Run on treg.to, 17 Aug 2026. Every figure below is from the Activity log of that run.
| Field | Value |
|---|---|
| Providers considered | 5 for keyword volume, 5 for Google results |
| Providers selected | serpstat.google.keywords.volume · dataforseo.google.serp.organic |
| Why | |
| Total cost of the run | $0.012 — 2 calls |
| Subscription cost avoided | $214/mo at list — Semrush $139 + SerpApi $75 |
| Time to completion | Under 15 seconds. The results call reported 10.4 s server-side |
| Data freshness | Google results timestamped 2026-08-17 02:06:40 UTC — live at call time. Keyword volume is the provider's monthly index |
What actually came back. 20 keywords submitted, 13 returned with data; seven had none, which is a real answer and worth more than an invented number.
| Keyword | Volume/mo | Difficulty | Intent |
|---|---|---|---|
| mcp server | 60,500 | 8 | informational |
| serp api | 12,100 | 17 | informational |
| agent skills | 8,100 | 5 | informational |
| reddit api | 5,400 | 9.65 | informational |
| tiktok api | 2,400 | 9 | informational |
| web scraping api | 1,900 | 23 | navigational |
| rank tracking api | 1,000 | 6 | informational |
| social media api | 320 | 8 | informational |
| backlink api | 210 | 3 | informational |
| keyword research api | 140 | 3 | informational |
| data enrichment api | 110 | 2 | informational |
| company data api | 70 | 5 | navigational |
| email finder api | 40 | 1 | informational |
And the live top 7 for serp api: serpapi.com, dataforseo.com, scrapfly.io, github.com, brightdata.com,
searchapi.io, you.com.
The honest read of this run:
mcp serverat 60,500/mo with difficulty 8 is the kind of finding the workflow exists to surface — but difficulty scores are a provider's model, not a fact. Treat the volume as data and the difficulty as an opinion you paid $0.0005 for.
Hand the agent 200 candidate titles. It returns real monthly volume, difficulty and intent for each, and flags where the current top three are weaker than you are. At $0.0005 a keyword that is a few cents, not a month of Semrush.
For any keyword, the agent pulls the live results, then pulls authority and referring-domain data for each page that ranks. You learn whether the leader is winning on links or on the page — which decides whether you write or go get links.
Run the same keyword set weekly. The agent keeps last week's positions and reports only what moved. You pay per run, so a paused project costs nothing while it is paused.
For one provider you already pay for, do — connect that key and those calls route through it, unmetered. The arithmetic changes at the second provider. This workflow touches keyword data, live results and backlinks; buying that separately is three accounts, three contracts and three API shapes to learn, for research you might run twice a month. treg.to is closer to OpenRouter for agent tools than to a data vendor: one base URL, one token, many providers behind it.
The credential is injected on the server. Your agent makes the real upstream request through treg.to's
/call/ endpoint; treg.to adds the key and relays the provider's answer back verbatim. No provider key
is ever written to your machine, your repo or your agent's context. Every call is recorded and attributed
to the token that made it.
Yes. Every endpoint has an id, and calling it by id calls that provider. If you want the choice made once
for the whole team, treg org pin <capability> --provider <provider> refuses calls to any other provider
of that capability.
(Worth doing here. Keyword databases disagree, so if you have been reporting Semrush
volumes to a client for a year, pin Semrush and keep the series consistent.)
Yes, and it takes precedence. Register the key once and every call to that provider routes through it — those calls are never metered against your treg.to balance. Your key always wins over treg.to's.
treg.to does not silently reroute you. That is deliberate: only you know which inputs you actually hold,
so treg.to relays your request rather than rewriting it. What your agent gets instead is the information
to recover — it already knows the alternatives and their parameters, so on a 429, a 5xx or a timeout it
can try the next provider and tell you which one it switched to. Failed calls are not billed. If a call
succeeded upstream but the answer was lost coming back, an Idempotency-Key returns the stored result
without paying twice.
Fractions of a cent for most SEO endpoints — Serpstat keyword volume at $0.0005 per keyword, DataForSEO organic results at $0.002 per call, Moz URL metrics at $0.006667, at the time of writing. The exact price is shown before the call, and treg.to adds no markup to what the provider charges. New teams start with $1.00 of free credit.
The installer registers treg.to's MCP server into Claude Code, Cursor and opencode automatically. Any MCP
client that supports the authorization spec can connect with OAuth. Anything that can run a shell command —
Codex included — can use the treg CLI or plain HTTP.
$1.00 of free credit on every new team. No credit card, no provider signup. treg.to is open source (AGPL).