Build your own SEO agent.

Claude Code, Codex or Cursor can already do keyword research, content plans and competitor checks for you. What they lack is real numbers. Give your agent keyword search volume, the current Google results and name checks through one plain HTTP API, with a hard daily budget so a looping agent cannot run up the bill.

Build one in 5 minutes

What an SEO agent needs, and where it gets it

The agent asks	Searchwire answers with
Which of these 300 keywords are worth writing for?	Monthly search volume, CPC and competition for all of them in one call: `/v1/keywords/volume`
Who ranks for this keyword today, and can we beat them?	The organic Google top 10 with rank, URL and title, and when it was observed: `/v1/serp`
Is this product or brand name free?	Domains across TLDs, GitHub and npm in one call: `/v1/names/check`
How much have I spent?	A `cost` on every response, and a daily cap per key: `/v1/account`

Every answer says where it came from and whether it was cached. Agents should never invent search volumes; when Searchwire cannot answer an item it returns an error for that item instead of a guess.

Build an SEO agent with Claude Code

Get a key. Buy $10 of credit above, or let your agent do it: POST https://searchwire.dev/v1/billing/checkout returns a payment link for you and a claim link for the agent.

Install the skill.

mkdir -p .claude/skills/searchwire
curl -s https://searchwire.dev/skill.md -o .claude/skills/searchwire/SKILL.md
export SEARCHWIRE_KEY=sw_...

Set a budget before letting it loop:

curl -s https://searchwire.dev/v1/account -H "Authorization: Bearer $SEARCHWIRE_KEY" \
  -H "content-type: application/json" -d '{"daily_budget_usd": 2}'

Ask it for keyword research. For example: "List 200 keywords our docs could rank for, get their US search volume from Searchwire in one call, keep the 20 best by volume and low competition, check who ranks for the top 5, and write a content plan."

The agent collects candidates, asks for all of them in one call (200 keywords cost about $0.05 fresh, less from the cache), narrows down, and only then looks at Google results for the few that matter. That order is what keeps research cheap.

Why not let the agent search Google itself?

Web search tools return pages, not numbers: no search volume, no competition, no reliable ranking positions. Scraping Google from an agent breaks quickly and is against Google's terms. SEO suites have the numbers but gate them behind monthly subscriptions. Searchwire is pay as you go, has no minimum beyond $10 of credit that never expires, and answers in JSON an agent can use directly.

Pricing

Fresh data costs what it costs us upstream plus 30%; answers from the shared cache cost 30% of that; domain, GitHub and npm checks are free. Typical: about $0.016 for one keyword's volume, about $0.03 for 100 keywords in one call. Details in the API docs.