Bluesky Trend Intelligence

TL;DR

Search public Bluesky posts through the official AT Protocol AppView, then receive normalized POST records plus one decision-ready SUMMARY. Every post includes authorHandle, text, createdAt, likeCount, repostCount, replyCount, quoteCount, engagementTotal, engagementVelocityPerHour, breakoutScore, fieldCompletenessScore, confidence, sourceUrl, and agentMarkdown. The summary adds ranked topics, author concentration, top breakout posts, source status, and query provenance.

Use it when a raw post export is not enough. It helps social teams spot accelerating conversations, researchers compare themes, agencies monitor client categories, founders watch market language, and agents consume evidence-linked records without guessing what an empty result means.

Run it in four ways:

1. Enter keywords or hashtags in the Apify Console.
2. Call the Actor through the Apify API from your application.
3. Schedule recurring searches to build a monitoring history.
4. Connect the dataset to webhooks, Make, Zapier, n8n, Sheets, BI tools, or an agent workflow.

If this Actor saves research time, leave a review. Reviews help prioritize the next buyer-facing improvements.

What does Bluesky Trend Intelligence do?

The Actor searches the public Bluesky network with the official app.bsky.feed.searchPosts endpoint. It does not need a Bluesky login, a proxy, a browser, or a buyer-supplied API key. It converts public AppView responses into stable, typed records designed for monitoring and downstream analysis.

A raw scraper tells you what posts exist. This Actor also explains what is moving:

• Engagement total combines likes, reposts, replies, and quotes as a number.
• Engagement velocity per hour divides observed engagement by post age, with a bounded floor for very new posts.
• Breakout score combines log-scaled engagement with velocity on a 0 to 100 scale.
• Topic clusters rank repeated meaningful words and preserve evidence record IDs.
• Author concentration shows which handles account for the most posts and engagement.
• Top breakouts provide direct public URLs for the strongest current signals.
• Source diagnostics distinguish a valid no-match from transport, HTTP, malformed JSON, and source-shape failures.

The scores are explainable signals, not predictions. Use the linked posts and normalized component fields when a decision needs human review.

Why monitor Bluesky now?

Bluesky conversations move quickly and often surface technical, media, creator, policy, startup, and community signals before they are reflected in slower reports. The public AT Protocol makes the source unusually integration-friendly, but teams still need normalization, deduplication, ranking, provenance, and honest failure behavior before the data is useful in a recurring workflow.

Common buyer questions include:

• Which posts around a category are gaining engagement fastest?
• Which terms repeatedly appear in the current result set?
• Are a few accounts dominating the conversation?
• What language should a content brief quote or respond to?
• Did a scheduled search return no matches, or did the source fail?
• Can the output go directly into an agent without a cleanup step?

This Actor answers those questions with typed fields and source links rather than an unstructured page dump.

What data do I get?

POST records

Each normalized post includes:

• schemaVersion: output contract version.
• recordType: always POST for post rows.
• id: stable SHA-256-derived record ID based on the AT Protocol URI.
• query: the buyer query that produced the row.
• platform: bluesky.
• source: the public AT Protocol AppView label.
• sourceUrl: a browser-readable public Bluesky post URL.
• uri and cid: AT Protocol identity and content address when exposed.
• retrievedAt, createdAt, and indexedAt: ISO 8601 timestamps.
• authorDid, authorHandle, and authorDisplayName.
• text and language.
• isReply: whether the post replies to another post.
• likeCount, repostCount, replyCount, and quoteCount.
• engagementTotal: numeric sum of the four engagement counters.
• engagementVelocityPerHour: age-normalized current engagement signal.
• breakoutScore: explainable 0 to 100 current breakout signal.
• fieldCompletenessScore: percentage of core fields present.
• confidence: bounded confidence based on completeness.
• agentMarkdown: a compact evidence block ready for retrieval or an agent prompt.

SUMMARY record

Every successful run ends with exactly one summary. It includes:

• Exact input queries and sort order.
• count and numeric mirror countNumeric.
• Number of unique authors.
• sourceStatus as ok or no_match.
• Ranked topics with counts, shares, and evidence IDs.
• topAuthors with post counts and total engagement.
• topBreakouts with scores, velocity, handles, and source URLs.
• errors: [] for a healthy completed source read.
• A query fingerprint for stable monitoring and audit joins.
• agentMarkdown with the leading topics and strongest breakout signals.

The same briefing is saved as AGENT_BRIEFING.md in the run key-value store.

How do I search Bluesky posts?

Enter one to ten queries. A query can be a phrase, keyword, brand, product category, person, hashtag, domain term, or Boolean-style expression supported by Bluesky search. Keep queries focused when you want interpretable topic shares.

Recommended inputs:

{
  "queries": ["artificial intelligence", "AI agents"],
  "sort": "top",
  "language": "en",
  "maxResultsPerQuery": 25,
  "minEngagement": 0,
  "includeReplies": false
}

For a recency-focused monitoring run:

{
  "queries": ["small business automation", "workflow automation"],
  "sort": "latest",
  "since": "2026-07-20T00:00:00Z",
  "language": "en",
  "maxResultsPerQuery": 50,
  "minEngagement": 2,
  "includeReplies": false
}

Input reference

Field	Type	Default	Purpose
queries	string array	required	One to ten public Bluesky searches.
sort	top or latest	top	Relevance/engagement or recency order.
since	ISO date or date-time	unset	Lower publication-time bound.
until	ISO date or date-time	unset	Upper publication-time bound.
language	BCP-47 string	en	Optional language filter passed to AppView.
maxResultsPerQuery	integer 1 to 100	25	Hard result cap per query.
minEngagement	nonnegative integer	0	Client-side minimum combined engagement.
includeReplies	boolean	false	Include or exclude reply posts.

The Actor rejects an invalid date range where since is later than until. Caps are enforced before output grows beyond the requested scope.

How is the breakout score calculated?

The score is intentionally explainable. It combines:

1. Log-scaled total engagement, so one huge account does not make every smaller signal invisible.
2. Engagement velocity per hour, which helps a newer active post compete with an older post that accumulated engagement slowly.
3. A hard 0 to 100 bound for predictable sorting and dashboards.

The underlying numbers remain in the record. A buyer can ignore breakoutScore and use engagementTotal, engagementVelocityPerHour, createdAt, and the direct source URL instead. Nothing depends on a hidden model or an opaque prompt.

This is not a forecast of future virality. It is a current signal computed from the public counters returned at retrieval time.

Worked example 1: find an accelerating AI conversation

Input:

{
  "queries": ["AI agents"],
  "sort": "latest",
  "maxResultsPerQuery": 25,
  "minEngagement": 1
}

Representative normalized output:

{
  "recordType": "POST",
  "query": "AI agents",
  "authorHandle": "builder.bsky.social",
  "text": "A practical agent needs tools, memory, and a clear approval boundary.",
  "engagementTotal": 38,
  "engagementVelocityPerHour": 12.667,
  "breakoutScore": 90.476,
  "fieldCompletenessScore": 100,
  "confidence": 0.96,
  "sourceUrl": "https://bsky.app/profile/builder.bsky.social/post/example"
}

Decision use: sort by velocity, open the linked evidence, then quote or brief only after reading the public post in context.

Worked example 2: compare category language

Input:

{
  "queries": ["customer support automation", "AI customer service"],
  "sort": "top",
  "maxResultsPerQuery": 40,
  "includeReplies": false
}

The summary might rank terms such as support, agent, workflow, ticket, and response. Each topic carries up to five evidence IDs. This supports a content brief without claiming that a repeated word equals customer demand.

Decision use: compare topic counts and shares, then inspect the highest-engagement evidence rows before changing messaging.

Worked example 3: monitor a brand mention

{
  "queries": ["examplebrand"],
  "sort": "latest",
  "since": "2026-07-21T00:00:00Z",
  "maxResultsPerQuery": 100,
  "includeReplies": true
}

Schedule the run hourly or daily. Store each dataset ID with its retrieval time. Use uri as the durable post identity and queryFingerprint to join comparable runs.

Decision use: alert only when a new URI appears or engagement velocity crosses your own threshold. This Actor does not send public replies or mutate Bluesky.

Worked example 4: identify concentrated voices

A summary can show that three authors account for most engagement in a result set. That prevents a team from mistaking a concentrated conversation for broad consensus.

Use topAuthors[].posts and topAuthors[].engagement together. A prolific low-engagement author and a single high-engagement author are different signals.

Decision use: diversify source review before presenting a market claim.

Worked example 5: build an agent research packet

Run several narrow queries, then feed only these fields into retrieval:

{
  "id": "stable-record-id",
  "query": "open source agents",
  "authorHandle": "researcher.bsky.social",
  "text": "Public post text",
  "createdAt": "2026-07-21T10:00:00.000Z",
  "engagementTotal": 18,
  "breakoutScore": 74.3,
  "sourceUrl": "https://bsky.app/profile/researcher.bsky.social/post/example",
  "agentMarkdown": "## @researcher.bsky.social ..."
}

Decision use: require the agent to cite sourceUrl and distinguish the author’s statement from its own interpretation.

Worked example 6: prove a valid no-match

A narrow query or time window can legitimately return no public posts. The Actor returns a SUMMARY with:

{
  "recordType": "SUMMARY",
  "count": 0,
  "countNumeric": 0,
  "uniqueAuthors": 0,
  "sourceStatus": "no_match",
  "topics": [],
  "topAuthors": [],
  "topBreakouts": [],
  "errors": []
}

That is different from a source outage. HTTP failures, malformed JSON, and payload-shape drift fail the run and write DEBUG_ZERO_PARSE.json with bounded diagnostic evidence.

How do scheduled monitoring workflows work?

Apify schedules can run this Actor at a recurring interval. Recommended monitoring pattern:

1. Keep the query set and sort mode stable.
2. Use since for the observation window you actually need.
3. Save the dataset ID and run timestamp.
4. Join rows by uri or stable id.
5. Compare new posts and engagement counters outside the Actor.
6. Route only evidence-backed changes to alerts, dashboards, or agents.

The Actor itself performs read-only collection. It does not like, repost, reply, follow, message, or change any Bluesky account.

Can I use the API and integrations?

Yes. The Actor works from the Console, Apify API, schedules, tasks, webhooks, and standard dataset exports. Output is available as JSON, CSV, XML, Excel, RSS, and the other formats supported by Apify datasets.

Useful integration patterns:

• Push new high-velocity rows into Slack or Teams through a webhook workflow.
• Append normalized records to Google Sheets for editorial review.
• Send datasets to BigQuery, Snowflake, PostgreSQL, or a warehouse pipeline.
• Trigger n8n, Make, or Zapier after a successful run.
• Feed agentMarkdown and source URLs into a retrieval or research agent.
• Build a lightweight category-monitoring API over saved datasets.

Why choose this Actor?

The leading general Bluesky scrapers are strong at broad extraction. This Actor is deliberately narrower and decision-focused.

Capability	Bluesky Trend Intelligence	Typical raw scraper
Official public AT Protocol source	Yes	Often
Login or proxy required	No	Usually no
Stable typed post schema	Yes	Varies
Engagement velocity per hour	Yes	Usually no
Explainable breakout score	Yes	Usually no
Ranked topic evidence IDs	Yes	Usually no
Author concentration summary	Yes	Usually no
Agent-ready Markdown per row	Yes	Usually no
Honest no-match contract	Yes	Varies
Debug artifact on source drift	Yes	Varies
Read-only behavior	Yes	Varies by product

Choose a broad scraper when you need followers, DMs, complete threads, or every general extraction mode in one run. Choose this Actor when you need search-driven public-post intelligence, monitoring records, source evidence, and explainable ranking.

Pricing

Launch pricing is pay per event:

• Actor start: $0.01.
• Dataset item: $0.003, equal to $3 per 1,000 emitted records before tier discounts.
• Buyer-controlled run cap: up to $5.
• Apify platform usage passthrough: off.

A dataset item is a normalized post or the run summary. The key-value-store briefing is not a second dataset charge. Small bounded searches stay inexpensive, while the run cap limits larger batches.

Pricing is positioned for intelligence output rather than the cheapest possible raw extraction. The listing exposes the score components and direct source links so buyers can judge the added value.

Reliability, empty results, and diagnostics

A successful run means the official AppView responded with valid JSON matching the expected post shape. Outcomes are explicit:

• sourceStatus: ok: one or more normalized posts passed filters.
• sourceStatus: no_match: the source was healthy and no posts passed the query and filters.
• Failed run: transport error, non-2xx HTTP status, malformed JSON, payload-shape change, or invalid input.

On a source or parser failure, the Actor writes a bounded DEBUG_ZERO_PARSE.json artifact with stage, message, URL, HTTP status when available, a limited response snippet, and retrieval time. It does not turn a blocked or malformed response into a billable-looking empty success.

Public network data can change or be removed. Engagement counts are observations at retrieval time. Deleted posts and moderation decisions remain controlled by Bluesky and the relevant Personal Data Server.

Tutorial: create a daily Bluesky category brief

1. Open the Actor and enter two to five focused category queries.
2. Choose latest when you care about current posts or top for the strongest result set.
3. Set since to the beginning of your monitoring window.
4. Keep maxResultsPerQuery between 25 and 100.
5. Set minEngagement only if low-interaction posts are noise for your use case.
6. Run once and inspect the SUMMARY plus direct source links.
7. Save the configuration as a task.
8. Add an Apify schedule.
9. Connect the dataset or webhook to your review destination.
10. Compare stable post IDs between runs before raising an alert.

A safe alert contains the query, post URL, observed counters, retrieval time, and the rule that triggered it. Avoid presenting a breakout score as a guaranteed future outcome.

FAQ

Does this need a Bluesky login?

No. It uses the public Bluesky AppView for public post search. It does not read private messages or private account data.

Does it use an unofficial scraper or browser automation?

No. The acquisition path is HTTP-only against the public AT Protocol AppView. That reduces cost, startup time, and anti-bot fragility.

Can it scrape followers or complete user feeds?

Not in this product. The Actor is optimized for search-driven trend and conversation intelligence. Use a broad profile or follower scraper when that is the required outcome.

Can it search hashtags?

Yes. Include the hashtag or term in queries. Search behavior ultimately follows the public Bluesky endpoint.

Can I include replies?

Yes. Set includeReplies to true. The default is false so top-level conversation signals are easier to interpret.

How is engagement calculated?

engagementTotal is likes plus reposts plus replies plus quotes. All component counters remain available separately.

Is the breakout score AI-generated?

No. It is deterministic math over engagement, post age, and a bounded logarithmic scale. The components are visible in every row.

Why can two queries return the same post only once?

The Actor deduplicates on the AT Protocol URI. The retained row preserves the query associated with the first accepted occurrence in the input order.

What does fieldCompletenessScore mean?

It measures presence of core identity, author, text, and time fields. It is a data-completeness indicator, not a truth score.

What happens if Bluesky changes its response shape?

The run fails closed and writes diagnostic evidence. It does not silently emit zero rows from an unexpected payload.

Can a no-match run succeed?

Yes. A healthy source with no matching posts returns an explicit zero-count summary with errors: [] and sourceStatus: no_match.

Can I use the output in an AI agent?

Yes. Use typed fields, agentMarkdown, query fingerprints, and direct source URLs. Require source citation and keep interpretations separate from observed post content.

Can I schedule monitoring?

Yes. Use Apify schedules or tasks. Keep comparable input stable and join rows by uri or id.

Does this Actor take actions on Bluesky?

No. It is read-only. It never posts, replies, likes, reposts, follows, or sends messages.

Is this affiliated with Bluesky?

No. This independent Actor uses publicly available AT Protocol data. Bluesky and AT Protocol names identify the data source. No endorsement or partnership is implied.

Legal and responsible use

Use the Actor only for lawful purposes and comply with applicable privacy, data protection, intellectual-property, employment, marketing, and platform rules. Public availability does not eliminate a buyer’s obligations. Minimize retained personal data, establish a lawful purpose, honor deletion or suppression duties where applicable, and avoid sensitive profiling or automated high-impact decisions without appropriate review.

Do not use output for harassment, stalking, discrimination, doxxing, credential discovery, or attempts to bypass platform controls. Do not infer protected traits from posts or profile metadata. The Actor collects public post search results and performs no account action.

Data may be revised, moderated, or deleted after retrieval. Preserve retrievedAt and sourceUrl, and re-check the source before making a consequential claim.

Support and feedback

Use the Actor Issues tab for reproducible bugs. Include the run ID, a redacted input, expected behavior, observed behavior, and whether the failure affected all queries or one query. Never post passwords, API tokens, private exports, or personal information that is not necessary to reproduce the issue.

Feature requests are welcome. The most useful requests describe the buyer decision, required fields, desired schedule, expected volume, and output destination.

Related Skootle Actors

Pair this Actor with:

• Google Trends Intelligence for search-interest direction, momentum, and related-query evidence.
• Website SEO Audit Intelligence for technical and content findings on sites discovered through social research.
• Website Tech Stack Intelligence for explainable technology fingerprints and lead qualification.

Use Bluesky Trend Intelligence for the conversation signal, then use a sibling Actor only when its separate data source supports the next decision.

Output contract summary

• One stable POST row per unique AT Protocol URI that passes filters.
• Exactly one SUMMARY row per healthy run.
• One AGENT_BRIEFING.md key-value record.
• ISO 8601 timestamps.
• Numeric engagement mirrors.
• Stable IDs and query fingerprint.
• Direct public provenance URLs.
• No generic untyped primary output.
• No silent success after source or parser failure.
• No external messages or platform mutations.