Domain Scout

Ask your coding agent for a domain name. It suggests names that fit your project, looks each one up with the registry or a registrar, and tells you which ones you can register, how certain that is, and the price when one is quoted.

A plugin for Claude Code and Codex. It runs on your machine, needs no API key, and requires Node.js 22 or newer.

What it looks like

A real run from 24 October 2026, trimmed:

You: Find a name and a .com for a language-learning app. The name has to work in English, Spanish and German.

The agent wrote 70 names in three waves. It kept to letters that all three languages pronounce the same way (no j, v, w, z, ch, ll or qu), and avoided names close to Duolingo, Babbel and Tandem. Every short Spanish word and invented name in its first wave came back unavailable, so it moved to blends and phrases; 8 of those were open at the standard price. Results change, so check again before you register.

Install

Claude Code

claude plugin marketplace add ossianravn/domain-scout
claude plugin install domain-scout@domain-scout-local

Start a new session and ask for a name, or run /domain-scout:find-domains.

Codex

codex plugin marketplace add ossianravn/domain-scout
codex plugin add domain-scout@domain-scout-local

Start a new session and begin with "Use Domain Scout to…", or pick find-domains from the skill list. In Codex, a plain naming request may not trigger the skill.

To try a local copy without installing it, run claude --plugin-dir /path/to/domain-scout. For Windows, WSL, credentials and troubleshooting, see the installation guide.

Things to ask

• "Find the shortest attractive domain for a short URL service."
• "Find a name for my second brain project. Explore metaphorical and invented names."
• "Find a domain for an app that helps teachers prepare lessons."
• "Find a name and a .com for an app that helps remote workers live anywhere for a month."
• "Check these exact domains and show registration and renewal prices: …"

If you haven't said what kind of name you want, the agent asks once, with examples for your project: descriptive, suggestive, metaphorical, playful, invented, a phrase across the dot, or a mix. It remembers your answer for the rest of the conversation. Lists of exact domains are checked straight away.

Reading the results

Every result keeps its source and check time. Prices keep their currency and period; when a source gives no renewal period, it stays unknown. Domain Scout doesn't buy domains, search resale listings or check trademarks.

How it works

%%{init: {"sequence": {"mirrorActors": false}}}%%
sequenceDiagram
    autonumber
    actor You
    participant Agent as Your agent<br>(Claude Code or Codex)
    participant Scout as Domain Scout<br>(local MCP server)
    participant Sources as Registries and<br>registrar API
    You->>Agent: "Find a .com for a language-learning app"
    Agent->>Scout: prepare_domain_search
    Scout-->>Agent: brief, open questions, naming directions
    opt Style not given yet
        Agent->>You: Descriptive, playful, invented, or a mix?
        You->>Agent: A mix
    end
    Agent->>Agent: Write candidate names
    Agent->>Scout: check_domains (one batch)
    Scout->>Sources: Exact lookups, paced per source
    Sources-->>Scout: Records, availability, prices
    Scout-->>Agent: Status, source, time and price per name
    Agent->>You: Shortlist with evidence

Your agent's own model writes the names, so there is no second model and no key to set up. The server only looks up exact domains. It sends out nothing but the domain names you check: to the registries and Vercel, and for the slow registries also to Cloudflare's and Google's public DNS. Your project description stays with your agent.

Both clients use the same skill and tools:

Where the answers come from

No source needs a key. Dedicated registry checks run before the general ones, and a source that doesn't sell an extension can't mark it taken. For the full routing order, see sources; for the bundled IANA routing data, see catalog.

Speed. Sources run in parallel, and repeat lookups within a minute come from a cache. A call waits up to 10 seconds by default: names not finished by then come back unknown and can be checked again. .dk, .nl and .be answer about one name per second. DNS screening answers for names already in use there in about a tenth of a second, so only open-looking names wait for the registry. Pass screen: false to ask the registry anyway.

Settings

All optional. Set them in the environment that launches Claude Code or Codex; .env files are not read.

Command line

The same lookups work without an agent:

node dist/cli.cjs check '{"domains":["example.com","example.dk"],"fresh":true}'
node dist/cli.cjs prepare '{"project":"A second brain app","namingStyles":["mixed"]}'
node dist/cli.cjs sources '{"tlds":["dk","com","nl"]}'

Pass - or nothing to read the JSON from stdin.

Development

npm ci
npm run check   # typecheck, build, offline tests
npm run smoke   # live lookups against public sources

npm run build regenerates dist/ and the skill's copy of the CLI. Commit them with your source changes, because Git installs run them as committed. Keep source files under 300 lines. Run npm run update:registry, then build, to embed newer IANA routing data.

More

• Installation guide: Windows, WSL, credentials, updates, troubleshooting
• Sources: routing order, limits and how each answer is read
• Naming: directions and shortest-domain search
• Catalog: the bundled IANA routing data
• Validation: what has been tested, and what hasn't

License

MIT. Bundled dependencies keep their own licenses; see third-party notices.