Firecrawl said on 20 November 2026 that Alexandria now covers more than four million podcast episodes, offering creators and research teams a way to search spoken discussions. Its Particle usage rules support on-demand podcast research, not a reusable transcript archive. One search, one full-transcript request and one episode-detail lookup total 35 credits at the listed rates, before other requests or AI costs.
The coverage figure is Firecrawl’s claim, not a guarantee that a particular show or episode is available. The usage documents below predate the November update.
What does a podcast lookup cost?
For a question that needs one complete episode, the published capability prices break down as follows:
Request | What it returns | Credits per call |
|---|---|---|
Matching transcript passages and compact episode records | 15 | |
The selected episode’s transcript | 15 | |
Episode information, with web and audio links when available | 5 |
Illustrative calculation: 15 + 15 + 5 = 35 credits. This assumes one search page and one selected episode, with no retries or additional providers. It excludes web searches, model or assistant fees, hosting, taxes and other account usage; it is not a measured bill.
If the search passages and surrounding dialogue answer the question without a full transcript, search plus episode details totals 20 credits. That retrieves less material, so it is not an equivalent replacement in every case. One search followed by transcript and detail requests for five episodes instead totals 15 + (5 × 20) = 115 credits, before extra search pages.
For a team example, five colleagues completing 20 of the 35-credit sequences each month would use 5 × 20 × 35 = 3,500 credits. That fits an otherwise unused 5,000-credit Hobby allowance, conditional on eligible Particle access. A new month-to-month Hobby subscription costs US$19; the US$16 figure is its annual-billing equivalent. Hobby top-ups add 1,000 credits per US$5 increment.
Do not assume a shorter transcript request costs less. The segment reference and full-transcript reference both list 15 credits, although the latter describes segments as cheaper. Check the live numeric price before budgeting a discount.
What can your app display?
The Particle schedule allows search, summaries and analysis in real-time query responses for authorized users. It restricts redistribution beyond those responses and uses that replace the original podcast.
Particle’s attribution rules require the supplied show name, episode title, creator or host, and original-episode link where provided. Preserve the delivered attribution; do not invent missing fields. Particle also does not warrant that content is authorized for a customer’s particular use. Technical access is not a blanket republication licence.
Can you keep transcripts for later questions?
Under Schedule B, stored podcast data cannot answer later queries, and agents cannot keep it in persistent memory. Model training, fine-tuning and validation are also prohibited.
The schedule lists narrow storage exceptions: transient operational caching, unchanged prior-response viewing while source access continues, and legally required retention. The stricter incorporated terms still control. An answer history is not permission to build a reusable transcript collection.
The same schedule requires takedown removal within 48 hours of notice, including histories. When source access ends, deletion is due within 30 days after Firecrawl’s notice, apart from legally required retention.
How do you check a quote and its source?
Search results do not include episode web or audio links. The five-credit episode-detail lookup may supply them, but neither is guaranteed. Budget for this step separately: finding relevant words and obtaining a route back to the recording are different operations.
Particle’s terms describe transcripts as machine-generated and potentially erroneous, and prohibit presenting them as authoritative verbatim records. Check important quotations against the original recording and surrounding discussion. If audio is unavailable, do not label the wording verified. A correctly transcribed guest’s statement still needs separate evidence if used as a factual claim.
If using segment transcripts, their timestamps start at the segment’s beginning. Convert that offset before linking to a moment in the full episode; otherwise the same number of seconds can point to a different passage.
Which terms and access rules apply?
There are two document dates to distinguish. Firecrawl’s B-1.0.0 schedule incorporates Particle text dated 23 July 2026; Particle’s live page says 21 October 2026. Neither date is the November coverage announcement.
Firecrawl’s general terms say the accepted document versions and hashes are recorded, material changes require renewed acceptance, and stricter incorporated provisions govern overlapping subjects. Check the team’s acceptance record and any amendments rather than assuming either public page settles every account’s rights. Resolve the storage and audio-use conditions before building around them.
For playback, Particle’s current terms distinguish publisher-source audio, fetched afresh for end users, from Particle-hosted audio for temporary internal processing. The episode reference does not establish which category every returned audio link belongs to. Confirm that distinction before adding a player.
Access is a separate check. Alexandria tools require an authenticated MCP connection; the keyless connection does not include them. Required provider terms go to an organization administrator for review. Agent runs skip providers with unaccepted terms, so a skipped source is not evidence that no relevant podcast exists.
The published Free plan offers 1,000 credits a month, but that balance alone does not establish Particle eligibility. The checked public documents do not settle a Particle-specific minimum plan, provider quota, regional availability or how every free Agent/MCP allowance affects these calls. For connection choices, see Firecrawl MCP account and usage limits.
Before committing, inspect the current tool contracts and execution prices and confirm access for the intended account. If the same app will combine podcast results with sales records, review Firecrawl company-data rights and costs separately: provider permissions are not interchangeable.
For an interview-preparation or research assistant, the useful starting point is a fresh question, enough context to interpret the answer, and a path back to the source. For a lasting transcript archive, the unresolved issue is permission to retain and reuse the material—not how many credits to buy.
