Playwright / Selenium Scrapers vs CVM Filings API
Teams often ask about Playwright Selenium vs CVM API after a portal scraper fails in CI. Browser automation can click through the CVM site; apicvm replaces that path with authenticated HTTP for companies and filings in the corpus.
Failure modes of browser scrapers
- DOM and anti-bot changes break selectors
- Headless fleets need proxies, captchas, and ops time
- Hard to map clicks to durable document IDs
- Parallelism hits rate limits and IP blocks
API path
resolve ticker → list documents → download file / extract textexport APICVM_KEY='apicvm_...'
export APICVM_URL='https://apicvm.dev'
curl -H "Authorization: Bearer $APICVM_KEY" \
"$APICVM_URL/v1/companies/resolve?query=PETR4&by=ticker"
curl -H "Authorization: Bearer $APICVM_KEY" \
"$APICVM_URL/v1/documents?ticker=PETR4&type=DFP&year=2024&perPage=20"When scrapers still appear
| Situation | Note |
|---|---|
| Document type not in apicvm corpus | Lacuna — scraper does not fix missing ingestion |
| One-off manual download | Portal UI is fine |
| Production product feature | Prefer API |
Related: Scraping CVM portal vs apicvm.
When to use each
| Use | Prefer |
|---|---|
| Stable ticker → PDF in an app | apicvm |
| Exploring a portal screen interactively | Browser |
| iPaaS HTML bots | Fragile — see Zapier scrapers vs API |
Migration checklist
- Replace “login + click download” with
GET /v1/documentsfilters - Persist
document.idinstead of temporary portal filenames - Move PDF fetch to
/v1/documents/:id/file - Delete headless Chrome from CI once parity is verified on your ticker set
If a type/year is missing, open an ingestion gap — do not reintroduce Selenium as the default path.
Current limitations
- apicvm only returns ingested documents — not “everything the portal HTML can see.”
- No claim of official CVM endorsement.
- Extraction remains async.
Next steps
Ready to integrate?
Get an API key and start querying Brazilian CVM filings programmatically.