Migrate from Context.dev

Most Context.dev endpoints have a direct Prefetch equivalent. The main changes are the base URL, the auth header, snake_case parameters, and four behaviors worth checking before you switch: caching defaults, crawl regexes, async crawls and the industry taxonomy.

The basics

	Context.dev	Prefetch
Base URL	`https://api.context.dev/v1`	`https://api.prefetch.io`
Auth header	`Authorization: Bearer …`	`X-API-Key: …`
Parameter style	camelCase (`mainContentOnly`), with some snake_case	snake_case (`only_main_content`), except `/brand` and `/enrich`
Formats	An object: `"formats": { "markdown": true }`	A list: `"formats": ["markdown"]`, or `formats=markdown,links` on `GET`
Cached pages	Scrapes accept copies up to 3 days old by default, crawls 1 day	Fetched live unless you set `max_age`
Industry	EIC industry and subindustry, inside the brand profile	IAB category and subcategory, from `/classify` or `/enrich`
SDKs	TypeScript, Python, Ruby, Go, PHP	None. Use any HTTP client, or generate one from the OpenAPI spec.

To keep Context.dev’s scrape caching, send max_age=259200000 (3 days, in milliseconds).

Endpoints

Context.dev	Prefetch
`POST /web/scrape`	`GET` or `POST /scrape`
`GET /web/urls`	`GET /map`
`POST /web/crawl`	`POST /crawl`, then `GET /crawl/{id}`
`POST /web/search`	`GET` or `POST /search`
`POST /web/answers`	`POST /answers`
`POST /brand/retrieve`	`GET /brand`, or `GET /enrich` for brand, company and industry together
`GET /brand/search`	`GET /resolve`
`GET /web/styleguide`	`GET /styleguide`
Logo Link (`logos.context.dev`)	Logo CDN (`logo.prefetch.io`)
`POST /batch/submit`	`POST /batch/scrape`, or `POST /crawl` for `"mode": "crawl"`
`POST /monitors` and its sub-paths	`POST /monitor` and its sub-paths
`POST /people/search`, `POST /people/enrich`, `POST /company/search`, `POST /news/search`, `POST /parse`	No equivalent

Prefetch also has endpoints with no Context.dev counterpart: /company, /classify with IAB categories, /screenshot, and batches of brand, enrich, company, classify, screenshot or styleguide.

Scrape

Context.dev	Prefetch	Notes
`url`	`url`
`formats: { markdown, html, screenshot, images, json }`	`formats: ["markdown", "html", "screenshot", "images", "json"]`	Prefetch also has `raw_html`, `links` and `summary`.
`jsonParams.schema`, `jsonParams.instructions`	`json_schema`, `json_prompt`
`sharedParams.mainContentOnly`	`only_main_content`
`sharedParams.includeSelectors`, `excludeSelectors`	`include_selectors`, `exclude_selectors`
`sharedParams.waitFor`	`wait_for` or `wait_for_selector`
`sharedParams.headers`	`headers`	Requests with headers are never cached on Prefetch.
`markdownParams.includeLinks`, `includeImages`	`include_links`, `include_images`
`maxAgeMs`	`max_age`	Milliseconds. Prefetch defaults to a live fetch.
`sharedParams.actions`, `profile`, `country`, `viewport`	No equivalent	Prefetch has `mobile` for a phone layout.
`formats.product`, `highlights`, `parse`, `bytes`	No equivalent	Use `json` with a schema for product fields.

Reading the response

What	Context.dev	Prefetch
Markdown	`markdown.data`	`data.markdown.content`
Did a format succeed?	`markdown.success` on every output	Formats you didn’t ask for are absent. A language model format that failed has `null` values and a reason in `data.warnings`.
Page title	`metadata.title`	`data.metadata.title`
Cache	`cache_metadata.status`	`data.metadata.cache_state`

Map

Context.dev	Prefetch
`domain=stripe.com`	`url=stripe.com`
`maxLinks`	`limit`
`search`	`search`
`includeSubdomains`	`include_subdomains`
`urlRegex`, `excludeUrlRegex`	No equivalent. Filter the returned `links` yourself, or use `search`.
`sitemapUrl`	No equivalent. Prefetch finds sitemaps through `robots.txt` and the usual locations. Use `sitemap=only` or `skip` to choose sources.
Response `urls[]`	Response `data.links[]`, each with `url`, `title`, `description`, `lastmod` and `source`

Crawl

Context.dev’s /web/crawl answers in the same request. Prefetch crawls are async jobs: POST /crawl returns an id, and you read pages from GET /crawl/{id} or receive them by webhook.

Context.dev	Prefetch	Notes
`maxPages` (default 100)	`limit` (default 25)	Both stop at 500.
`maxDepth`	`max_depth`
`urlRegex`, `excludeUrlRegex`	`include_paths`, `exclude_paths`	Context.dev matches the full URL (`^https?://[^/]+/blog/`). Prefetch matches the path and query (`^/blog/`), so drop the scheme and host from your patterns. Prefetch takes a list of patterns.
`followSubdomains`	`allow_subdomains`
`useMainContentOnly`	`scrape_options.only_main_content`
`includeSelectors`, `excludeSelectors`	`scrape_options.include_selectors`, `exclude_selectors`
`waitForMs`	`scrape_options.wait_for`
`maxAgeMs`	`scrape_options.max_age`
Response `results[]`	`data.data[]` on `GET /crawl/{id}`, with `next` for the following page

Search

Context.dev	Prefetch
`query`	`query`
`numResults` (1 to 100)	`limit`
`includeDomains`, `excludeDomains`	`include_domains`, `exclude_domains`
`freshness: "last_24_hours"`, `"last_week"`, `"last_month"`, `"last_year"`	`time_range: "day"`, `"week"`, `"month"`, `"year"`. Prefetch also has `"hour"`.
`country`, `languages`	`country`, `language` (two-letter codes)
`markdownOptions`	`scrape_options: { "formats": ["markdown"] }`

Answers

The request is almost the same:

Context.dev	Prefetch
`task`	`task`
`json_format`	`json_format`
`mode: "fast"` or `"ultra"`	`mode: "fast"` or `"ultra"`. `ultra` is the REST default on both.
`timeoutOpts`	`timeout_opts`
`tags`	`tags`

See Answers.

Brand

Context.dev’s POST /brand/retrieve takes a type. Each type maps to a different Prefetch call:

Context.dev `type`	Prefetch
`by_domain` with `domain`	`GET /brand?url={domain}`, or `GET /enrich?url={domain}`
`by_url` with `direct_url`	`GET /brand?url={page URL}`
`by_name` with `name`	`GET /resolve?name={name}`, then `/brand` with the domain
`by_email` with `email`	Take the domain from the address, then `/enrich`. See Enrich signups.
`by_transaction` with `transaction_info`, `mcc`, `city`, `country_gl`	`POST /transaction` with `descriptor`, `mcc` and `country`
`by_ticker` with `ticker`	No equivalent. `/classify` returns a domain’s tickers, but doesn’t look a ticker up.

Brand fields

Context.dev `brand.`	Prefetch `data.`
`title`	`company.display_name` (from `/company` or `/enrich`)
`description`, `slogan`	`classification.description`, `classification.tagline_description`
`colors[]`	`brand.colors.brand_colors[]`
`logos[]`, with `type` and `mode`	`brand.logos.primary`, `brand.logos.secondary[]`, `favicon_url`, `icon_url`
`socials[]`, a list of `{ type, url }`	`brand.social_links` (or `company.social_links` in `/enrich`), an object keyed by platform
`address`	`company.addresses[]`
`email`, `phone`	`company.emails[]`. There’s no phone field.
`links`	`company.links`
`stock`	`classification.stock_tickers[]`
`employees`	`company.employeeCount`
`industries.eic[]`	`classification.category` and `subcategory`, in the IAB taxonomy
`is_nsfw`	`company.isNsfw`

Styleguide

Context.dev	Prefetch
`domain=stripe.com` or `directUrl`	`url`
`colorScheme=dark`	`dark_mode=true`
`maxAgeMs`	Always read live. A repeat request can come from the request cache.

Prefetch’s styleguide also returns exports.css and exports.tailwind, ready to paste. See Styleguide.

Logo Link

Context.dev	Prefetch
`https://logos.context.dev/?publicClientId=brandLL_…&domain=stripe.com`	`https://logo.prefetch.io/stripe.com?token=pk_…`
`type=icon`	The default. Prefetch serves one square logo per domain.
`type=wordmark`	No equivalent
`theme=light` or `dark`	No equivalent. Prefetch’s `theme` only sets the background a JPEG is flattened onto.
A monogram when no logo is known	A letter mark by default. `fallback=404` or `transparent` to change it.

Prefetch adds size (16 to 512 pixels), format (png, webp, jpeg) and greyscale. Get your token from GET /logo/token.

Batches

Context.dev	Prefetch
`input.mode: "scrape"` with `input.data.urls[]`	`POST /batch/scrape` with `urls[]`, up to 500 per batch
`input.mode: "crawl"`	`POST /crawl`
`input.data.formats`, `sharedParams`, `jsonParams`	`scrape_options`, with the same fields `/scrape` takes
Per-URL `itemId` and `meta`	No equivalent. Results carry their `url`. Key your records on it.
`webhook` or `webhookUrl`	`webhook`
`GET /batch/{id}/results`	`GET /batch/scrape/{id}`, following `next`

Context.dev reserves credits when you submit and refunds what’s unused. Prefetch charges results as they complete, and never for failed pages.

Monitors

Context.dev	Prefetch
`target: { type: "page", url }`	`targets: [{ type: "scrape", urls: [url] }]`. One monitor can hold up to 50 targets.
`target: { type: "page", instructions }`	A `scrape` target plus a `goal` on the monitor
`target: { type: "extract", schema, instructions }`	A `scrape` target with `change_tracking: { modes: ["json"], json_schema, json_prompt }`
`target: { type: "sitemap" }`	A `crawl` target, which reports pages that appear and disappear
`target: { type: "search", query }`	`targets: [{ type: "search", queries: [query] }]`
`include_selectors`, `exclude_selectors`	`scrape_options.include_selectors` and `exclude_selectors` on the target
`schedule: { type: "interval", frequency: 6, unit: "hours" }`	`schedule: { text: "every 6 hours" }`, or `cron` with a `timezone`
Runs and changes	Checks, and the pages in each check marked `changed`, `new` or `removed`
`POST /monitors/{id}/run`	`POST /monitor/{id}/run`
Webhooks	`webhook` and email alerts. See Alerts.

See Monitors for targets, schedules and judging.

Webhooks

Both sign deliveries the same way: t=<unix seconds>,v1=<hex HMAC-SHA256 of "<t>.<raw body>">. Change the header you read from X-Context-Signature to X-Prefetch-Signature, and use the secret Prefetch returns when you create the job or monitor. See Verify signatures.

Errors

Both return a machine-readable error_code. Prefetch adds retryable, so you can decide whether to retry without a list of codes. A request over your credit balance returns 403 with CREDITS_EXHAUSTED on Prefetch, where Context.dev returns 402. See Error handling.

Checklist

Swap the base URL and the auth header.
Turn formats objects into lists, and rename parameters to snake_case.
Set max_age wherever you relied on Context.dev’s default caching.
Rewrite crawl regexes to match paths, and poll or webhook for crawl results.
Map EIC industries to IAB categories where you store them.
Update your webhook signature header and secrets.

No-code tools

Compare

Migrate

Migrate from Context.dev

The basics

Endpoints

Scrape

Reading the response

Map

Crawl

Search

Answers

Brand

Brand fields

Styleguide

Logo Link

Batches

Monitors

Webhooks

Errors

Checklist

​The basics

​Endpoints

​Scrape

​Reading the response

​Map

​Crawl

​Search

​Answers

​Brand

​Brand fields

​Styleguide

​Logo Link

​Batches

​Monitors

​Webhooks

​Errors

​Checklist

The basics

Endpoints

Scrape

Reading the response

Map

Crawl

Search

Answers

Brand

Brand fields

Styleguide

Logo Link

Batches

Monitors

Webhooks

Errors

Checklist