Your agent needs to know what people are really saying about something. Not what the model picked up in training, and not a summary of a summary. The posts themselves.
The official way in is X’s own API. You create a developer account, get API keys, and connect them to your agent. Your agent can’t do any of that by itself, so the task stops and waits for you. The unofficial ways keep breaking.
So the question gets skipped, and your agent keeps working from what it saw in training.
Through TaskFuel, it reads the posts for half a cent, with nothing to set up for X.
Read a whole timeline
The search comes from Otto, one of the providers in the TaskFuel catalog. Here it is pulling Andrej Karpathy’s posts:
→ GET /tweet-search?query=from:karpathy
20 posts · most recent first
paid $0.005You get the full text of every post, long ones included, plus likes, retweets, replies, quotes and when it was posted. The longest of the twenty was about 1,600 characters. In it, Karpathy gave Opus 5 the first paragraph of The Lord of the Rings and about $10 of tokens. The whole post came back, down to the last sentence, with its 28,704 likes and 1,692 replies.
This matters more than it sounds. A lot of cheap scrapers cut long posts off partway, so your agent gets the start of the argument without the end. If it summarizes those, it sounds sure of itself and gets it wrong, and nothing in its answer tells you.
Or search for a topic
You’re not limited to one person. You can search by keyword, by $CASHTAG, or
by @handle to find mentions, and combine terms with OR. That way your agent
can look up a topic instead of a person.
We searched for people describing things their agent bought on its own:
→ GET /tweet-search?query="my agent bought" OR "agent paid for it itself"
20 posts
paid $0.005@racheleizner (30 likes)
my agent bought me my now favorite shoes on sale last month & i
had a really hard time keeping m…
@homerter85
my agent bought the dip for me while i was asleep last night.
woke up to a notification, position al…
@visionandvibehq
Yeah, I know, but my agent bought a radio and is out of mone…Nobody wrote those for a survey. That’s why your agent should read the posts, not someone’s summary of them. Both searches together cost one cent.
When this is useful
It works best as one step inside a bigger task. Your agent is writing a launch post and checks how people already use a phrase. It’s sorting support tickets and looks for anyone else who hit the same bug this week. It’s about to recommend a tool and reads what users said about it last month, instead of trusting the tool’s own website.
None of those is worth stopping to set up an X developer account. Each one is worth half a cent, and your agent can do it right away.
Two searches. One cent.
Both prices above are what we were actually charged. Your agent sees the price before it pays, so it can skip a search that isn’t worth half a cent.
No X developer account, no API keys, no integration code to maintain. Searches are paid from the same balance your agent uses for everything else in the catalog.
See the X/Twitter search tool for setup and a dated price check. Connect your agent and the first $5 is on us.
Search X for the latest posts about this product. Read the full text of each post, not the preview, and tell me what people keep bringing up, with a link to one post for each point. Show me what the search cost.
