Structured data reference

research, inspect, and execute provide progressive access to Firecrawl Alexandria’s data operations. Discovery and schema inspection are free; execution is paid. Both the CLI and Origo MCP use the same underlying engine.

Discovery modes and filters

webctx research "podcast conversations" --limit 5
webctx research --view sources --limit 5
webctx research --view groups --limit 5
webctx research --mode catalogue --view operations --sources particle --limit 10 --offset 0
webctx research --mode catalogue --view operations --query "podcast conversations" --include inputs,output,examples

research supports query, urls, sources, categories, groups, operations, view, include, limit, offset, mode, and raw. Views are sources, groups, and operations; include controls are inputs, output, and examples.

• Ranked discovery finds relevant operations by meaning. It has no native offset pagination.
• Catalogue mode browses the pageable inventory. For a filtered source with over 100 operations, local relevance ranking applies only within the fetched page. The response flags incomplete ranking and provides the next page.
• Exact operation IDs take precedence over descriptive filters.

Responses include machine-readable next-step arguments; catalogue browsing can include the upstream native_request for traceability.

Inspect an operation

webctx inspect particle/podcasts/episodes/search

An inspection returns upstream field names, types, required/optional inputs, allowed values, defaults, response shape, price, and examples when available. It also provides a suggested next call:

{
  "next": {
    "tool": "execute",
    "arguments": {
      "calls": [{
        "id": "particle/podcasts/episodes/search",
        "inputs": {"semantic_search": null, "limit": 25}
      }]
    }
  },
  "fill_before_execution": ["semantic_search"]
}

Illustrative shortened response. Required inputs with null are placeholders, not runnable values. If a source has no published examples, webctx may generate clearly labeled illustrative shapes without claiming they came from the provider.

Use webctx inspect particle to browse a source’s operations. Add --raw to see the upstream response in full.

Execution and pagination

webctx execute particle/podcasts/episodes/search --inputs '{"semantic_search":"AI agents","limit":2}'

To run up to ten operations in one call:

webctx execute --calls @requests.json

Each operation executes independently. One provider’s terms or validation rejection does not need to discard other successful items. Results contain native data, operation IDs, provider receipt information, and estimated/reported credits. Firecrawl charges are distinguished from provider-specific data fields also named credits.

When the provider supplies a continuation cursor, offset, page token, search_after, or continuation link, WebCTX provides a copyable next call. Use a new request ID for each new page or changed query.

Credit limits and safety

The hosted Origo server enforces a default 200-credit per-request ceiling before paid execution. It validates the supplied inputs and requires bounds for per-record operations when necessary. There is no user-facing argument to raise the hosted ceiling. A deployment may configure a lower limit.

The response can show estimated_credits and provider-reported credits_used. Estimates are not audited billing statements. Provider-account credits and third-party charges remain authoritative.

If a source requires acceptance of third-party terms, the response supplies the upstream terms URL. WebCTX does not accept terms or create provider accounts automatically.

Retries and idempotency

Paid requests have an idempotency request_id. If an execution has an uncertain network result, retry the same inputs with the same ID. Do not use that ID for another query or another page. WebCTX does not automatically replay a potentially charged operation.

Within one warm Origo worker, identical successful/partial replays can be served from a 15-minute in-memory cache. Such a replay reports replayed: true and credits_charged_this_request: 0. The cache prevents repeated upstream execution in that worker, but does not prove that a replay on another worker avoided new charges. Concurrent uses of one ID are serialized within a worker.

Mixed-operation batches return successful items alongside failures. After resolving a failed provider’s access, execute only the failed operation with a new request ID instead of resubmitting already successful paid items.

Rate-limit errors may provide retry_after_seconds, retry_at, and retryable fields. Free catalogue lookups can use short warm-worker caches and local throttling, but upstream quotas still apply.

Advanced output

--raw or "raw": true requests the complete upstream Firecrawl response instead of the normal concise projection. It may contain extra fields and provider-specific detail.

The MCP readOnlyHint metadata is not a promise that an operation is free: execute is still capable of incurring paid credits.

See Research structured data, CLI reference, and Origo technical reference.