# Discover (/docs/engines/discover) Search many platforms for a topic, keep only posts from your dates, score each against your question and get one de-duplicated list Discover searches several platforms for a topic and keeps only posts from your dates. It scores each post against the question you ask and removes duplicates. Use it when you need depth and a verdict on every row, not just a list of hits. A dry run shows the plan and the price for 0 credits. ## Price it for free Put the topic words in `query` and your question in `context`. `depth=quick` keeps the search to a few fast platforms. Add `dry_run=true`. ```bash title="cURL" curl -G "https://www.socialcrawl.dev/v1/prism/discover" \ -H "x-api-key: YOUR_API_KEY" \ --data-urlencode "query=electric cars" \ --data-urlencode "context=What do women over 50 say about owning an electric car?" \ --data-urlencode "depth=quick" \ --data-urlencode "period=90d" \ --data-urlencode "dry_run=true" ``` ```json title="Response (trimmed)" { "success": true, "data": { "valid": true, "estimate": { "credits": { "min": 3, "expected": 8, "max": 25 }, "by_stage": { "collect": { "expected": 3 }, "judge": { "expected": 5, "max": 22 } }, "answers_in_call": true }, "plan": [ { "platform": "youtube", "search": "electric cars" }, { "platform": "tiktok", "search": "electric cars" }, { "platform": "news", "search": "electric cars" } ], "suggested_max_credits": 25, "next_step": "Send the same request without dry_run to run it; it answers in this call and is expected to cost about 8 credits (at most 25)." }, "credits_used": 0 } ``` `plan` lists each search the run will make. `by_stage` splits the price: `collect` is the platform searches and `judge` is scoring each row against your `context`. `answers_in_call: true` means the rows come back in the same response. ## Run it Send the same request without `dry_run`. ```bash title="cURL" curl -G "https://www.socialcrawl.dev/v1/prism/discover" \ -H "x-api-key: YOUR_API_KEY" \ --data-urlencode "query=electric cars" \ --data-urlencode "context=What do women over 50 say about owning an electric car?" \ --data-urlencode "depth=quick" \ --data-urlencode "period=90d" ``` ```json title="Response (trimmed)" { "success": true, "data": { "items": [ { "platform": "news", "record_type": "article", "url": "https://www.motortrend.com/reviews/2025-acura-adx-yearlong-review-update-6-ev-switch-gas-pros-cons", "published_at": "2026-08-05T07:00:00.000Z", "in_window": true, "judgments": { "relevance": { "p": 0.94 } }, "matched_keywords": ["electric cars"] } ], "run": { "run_id": "run_EpfL4dXi6qpYZCBRNiybdK", "status": "completed", "credits": { "spent": 8, "by_stage": { "collect": 3, "judge": 5 } } } }, "credits_used": 8 } ``` Every request is a run with a `run_id`. A small request answers in the call. A large one answers `202` and continues in the background, and you follow it at `data.next.status_url`. `judgments.relevance.p` is the score from 0 to 1. ## What does it cost? You pay each platform's own price for a search, and only when it returns rows. A search that finds nothing, fails or comes from cache costs 0. Scoring rows against your keywords is free. Scoring against a `context` that differs from the keywords costs 1 credit per started 15 rows. `max_credits` is the most a run may spend. A GET that sends none may spend up to 500. A request whose expected cost is above `max_credits` is refused free. ## How do you follow a topic? Send a `POST` with a `schedule` and the request runs again by itself, delivering only rows it has not delivered before. Price the schedule with `dry_run` first. Events for each run arrive at an event webhook or from `GET /v1/events?types=discover.*`. Both are free. ## Next steps - [Discover reference](/docs/prism/discover.md): Every view, filter, schedule and price, with real requests to copy. - [Find creators](/docs/engines/find-creators.md): Rank the creators behind a topic instead of its posts. - [Watch something over time](/docs/webhooks-and-events.md): Schedules, event webhooks and the feed.