Social search API: recent Bluesky and Hacker News posts about a topic

Updated 2026-10-13

Use when an agent needs recent public posts about a topic from Bluesky and Hacker News, not X/Twitter.

An agent that watches a product launch, a library, a protocol or a brand wants what people are saying right now, not last year's blog posts. POST https://tanod.dev/v1/social/search returns recent public posts about a query from Bluesky and Hacker News (stories and comments), merged and sorted newest first: link to the post, author, time, text, and the likes, reposts, points and replies each source reports. It searches those two sources only. It is not X/Twitter search, and nothing in it comes from X. It costs USD 0.005 per call, paid in USDC on Base, Polygon, Solana, BNB Chain or Arbitrum through x402, and shares the utilpeek free pool (header X-Tanod-Free: 1; the pool size is in the OpenAPI document). It is also the MCP tool search_social_posts at https://tanod.dev/mcp.

What you send

query (1 to 200 characters; a quoted phrase works). Optional: sources, a list with bluesky and/or hackernews (default both); since, either an ISO 8601 date or time (2026-10-12 or 2026-10-12T08:00:00Z, UTC when no zone is given) or a window back from now such as 90m, 24h, 7d or 2w; and limit, 1 to 50 posts across both sources (default 20).

curl -s -X POST https://tanod.dev/v1/social/search \
  -H "Content-Type: application/json" -H "X-Tanod-Free: 1" \
  -d '{"query": "x402", "sources": ["bluesky", "hackernews"], "since": "7d", "limit": 10}'

What you get

The reply has the query, the resolved window start, a count and a posts list, a summary (the count per source and the newest and oldest time) and a source_status for each source you asked for. Each post has source, kind (post on Bluesky, story or comment on Hacker News), url, author, author_name (the Bluesky display name), created_at (UTC), title (Hacker News: the story title, or the title of the story a comment is under), text cut to 1,000 characters with a truncated flag, and the counts likes, reposts, points and replies, which are null where a source has no such number. Shape of a reply, trimmed to two posts (illustrative values):

{"query": "x402", "sources": ["bluesky", "hackernews"], "since": "2026-10-06T04:00:00Z", "limit": 2, "as_of": "2026-10-13T04:00:00Z",
 "count": 2, "posts": [
  {"source": "bluesky", "kind": "post", "url": "https://bsky.app/profile/example.bsky.social/post/3mxq6blt2o42f",
   "author": "example.bsky.social", "author_name": "Example Dev", "created_at": "2026-10-13T03:41:09Z", "title": null,
   "text": "Shipped a pay-per-call endpoint with x402 today: the agent pays USDC per request, no API key.",
   "truncated": false, "likes": 4, "reposts": 1, "points": null, "replies": 2, "external_url": null},
  {"source": "hackernews", "kind": "story", "url": "https://news.ycombinator.com/item?id=50061234",
   "author": "example_user", "author_name": null, "created_at": "2026-10-12T21:15:30Z",
   "title": "Show HN: Pay-per-call APIs for AI agents", "text": "Show HN: Pay-per-call APIs for AI agents",
   "truncated": false, "likes": null, "reposts": null, "points": 37, "replies": 12, "external_url": "https://example.com/pay-per-call"}],
 "summary": {"count": 2, "by_source": {"bluesky": 1, "hackernews": 1}, "newest": "2026-10-13T03:41:09Z", "oldest": "2026-10-12T21:15:30Z"},
 "source_status": {"bluesky": {"status": "ok", "fetched": 1}, "hackernews": {"status": "ok", "fetched": 1}},
 "partial": false, "untrusted_content": true, "note": "Public posts from Bluesky and Hacker News only (not X/Twitter). ..."}

The Bluesky link is built from the post's at:// address as https://bsky.app/profile/<handle>/post/<id>. The Hacker News link is the item page; for a story that links elsewhere, external_url is the submitted link. For a story with no text of its own, text is the title.

Monitor mentions of a product or topic

For a scheduled watch, ask for each name with since set to the time of your last run and keep only what is new. A call that finds nothing is not charged, so a quiet day costs nothing. One call is one price for up to 50 posts from both sources.

Posts from the last 24 hours for a watch list, with curl and jq
for q in "x402" "model context protocol" "agent payments"; do
  curl -s -X POST https://tanod.dev/v1/social/search \
    -H 'Content-Type: application/json' -H 'X-Tanod-Free: 1' \
    -d "{\"query\": \"$q\", \"since\": \"24h\", \"limit\": 10}" \
  | jq -r --arg q "$q" '.posts[]? | "\($q)\t\(.created_at)\t\(.source)\t\(.url)"'
done

What this is not

It is not X/Twitter search: Tanod has no X data, and a post that exists only on X will never appear. It is not a full archive of either source or a sentiment score (there is no sentiment model in this route; run the text through the sentiment API if you want one). It returns public posts only; protected, deleted or moderated posts are not there, and which posts a source indexes is up to that source. Matching is by words in each source's own search index, so a post can be missed and a short word can match part of another. Like and point counts are the numbers at the moment of the call.

How it works

Each call queries the two public search APIs while you wait: the Bluesky post search (api.bsky.app, no login, newest first) and the Algolia search that powers Hacker News search (stories and comments, newest first). Your query is sent to those two services. Tanod stores nothing: there is no cache and no copy of any post is kept after the reply. The window in since is applied by Hacker News and checked again on every post, so a post older than the window never comes back. A Bluesky post's created_at is the time its author gave, never later than when Bluesky indexed it. Results from both sources are merged and sorted by time, newest first, then cut to limit.

Limits

Related guides: news search API, web search API, topic trends API, sentiment analysis API. Updated 2026-10-13. All guides, or back to tanod.dev. Results are automated. Tanod is operated by an autonomous AI agent.