Every webctx command, common flags, and output format.

Commands

webctx --help
webctx --version

# Web research
webctx search <query> [--exclude domain1,domain2] [--keyword "phrase"]
webctx read-link <url> [--question "what to find"]
webctx map-site <url> [--query topic] [--limit N] [--offset N] [--path-prefix /api] [--language en|all]

# Structured data
webctx research [question] [--mode ranked|catalogue] [--view sources|groups|operations] [--sources IDs] [--include inputs,output,examples] [--limit N] [--offset N]
webctx inspect <operation-or-source-id> [--raw]
webctx execute <operation-id> --inputs '{"field":"value"}' [--request-id ID]
webctx execute --calls '[{"id":"source/operation","inputs":{}}]'

# Configuration
webctx status
webctx providers [status|enable <brave|tavily|exa>|disable <brave|tavily|exa>]
webctx keys list
webctx keys add <brave|tavily|exa|firecrawl> --stdin
webctx keys remove <brave|tavily|exa|firecrawl> <managed-key-index>

Use quotes around URLs containing #, ?, or shell-special characters.

Flag Behavior
--exclude domains Removes specified domains, separated by commas, in addition to built-in exclusions.
--keyword phrase Prefers Exa for a quoted phrase, with Brave/Tavily fallback. It is not a verified full-text filter.

Normal search uses available enabled providers in parallel, removes duplicate URLs, and returns up to 35 ranked links with provider snippets. It does not automatically open result pages.

Enable or disable providers with webctx providers enable exa and webctx providers disable exa. This preference is separate from whether you have configured keys.

Flag Behavior
--question "..." Selects relevant sections and can explore a bounded set of related documentation links.

Without --question, the response follows the specified URL. Native GitHub URLs keep source ranges, comments, review threads, jobs, and other selectors when possible. See GitHub URL reference.

map-site

Flag Default Behavior
--query None Prioritizes pages related to the topic.
--limit 30 Number of returned pages (1–100).
--offset 0 Zero-based result offset.
--path-prefix None Includes only paths under the given prefix.
--language en Locale filter; use all to include translated pages.

Mapping can discover at most 5,000 URLs from the configured provider request. Results include total counts and a copyable continuation when more pages are available. Explicit --limit 0 is invalid; omit the flag for the default.

Structured-data commands

research can search by natural-language question, or use --mode catalogue to browse a paginated list. Views are sources, groups, and operations. Additional filters include --urls, --sources, --categories, --groups, and --operations. Use --include inputs,output,examples to expand specific contract fields.

inspect accepts an operation or source ID. Use --raw for the full upstream contract.

execute accepts an operation ID and a JSON object of inputs, or --calls for a batch of up to ten operations. Prefix a JSON filename with @ when arguments are too long:

webctx execute --calls @requests.json
webctx execute particle/podcasts/episodes/search --inputs @inputs.json

--request-id supplies an idempotency ID for retries of the same paid execution. Do not reuse it for a new query or page. For prices, credit limits, return fields, and retries, see Structured data reference.

Credentials and settings

webctx status displays search provider enablement and locally configured key counts, not live credit balances. keys list also reports managed macOS Keychain slots. keys add reads one key from standard input and keys remove removes a numbered managed key; they never print key values.

Search provider settings are stored in your OS user config directory under webctx/config.json, or at WEBCTX_CONFIG_FILE if set.

See Credentials for environment variables and Rate limits and concurrency for scheduling overrides and state locations.

Output formats

Command Output
search Markdown with titles, URLs, and short provider snippets.
read-link Markdown/text with source attribution and caveats.
map-site Site URL inventory with counts and pagination.
research JSON with sources/operations and next-step arguments.
inspect JSON with contract, price, and suggested execution.
execute Provider data and per-operation results, credits, and continuations.

Raw source files, diffs, and patches remain close to their upstream representation. Errors include provider-specific limitations when known.