My biggest takeaway from SF Tech Week: the center of gravity has already moved. Last year everyone was talking about agents. This year, a lot of the interesting conversations were about everything agents need once they actually start working: routing, search, data, evals, observability, latency, inference cost, GPUs, deployment, reliability. The question used to be: Can an agent do this? Now it increasingly feels like: How do we make a million agents do this reliably, cheaply, and controllably? That shift matters. Companies like Tavily are a good example. Most end users may never know the name, but a huge number of agent workflows need search and web retrieval underneath. Once agents become real products, the boring-looking infrastructure around them becomes extremely valuable. Another thing Tech Week messes with is your perception of startup outcomes. You meet founder after founder who raised, got acquired, or somehow went from starting a company to becoming a VP / GM / Head of X at a much larger company in a few years. After hearing enough of these stories, your brain starts treating this as normal. It obviously isn't. Tech Week is basically the Instagram feed of the startup distribution. You are meeting people from the far right tail. The companies that ran out of money, never found PMF, or quietly dissolved are not usually on stage telling you their story. And even we got acquired can mean very different things: a great strategic outcome, an acquihire, investors getting some money back, or simply a soft landing. Startups don't just increase expected upside. They massively increase variance. One other thing I kept hearing was FDE: Forward Deployed Engineer. The role is interesting because it blurs boundaries that used to feel much cleaner. Engineer builds. Sales sells. PM translates. But more and more, the valuable person seems to be someone who can talk to the customer, understand the business problem, talk to engineers, turn ambiguity into something executable, and actually ship it. Engineer + Product + Consultant + Sales. I increasingly think this is less a job title than a survival skill. AI is already crawling into every industry and every workflow. So maybe the interesting question is no longer what AI can do. It is: What workflow has not been rebuilt yet? And after spending a few days around all of this, I came away with two contradictory feelings: How is everyone here doing something this ambitious? And: Maybe I should be doing more too. The Bay Area has never had a shortage of ideas or smart people. The scarce things are judgment about what is worth doing, the willingness to stay with it for years, and the ability to get other people to care.
Tavily on X
The largest group of people says Tavily quietly does the search job inside their agents and apps; the argument around it is price, and how it scores and bills next to Exa and DeepSeek's new, far cheaper Search API.
First written 27 Oct · what changed · from 93 posts by 77 people
Tracking posts that say “tavily”
Track this for your marketPosts per day, last 30 days
as of Nov 3, 11:01 UTCWhat changed
as of 27 Oct · checked once a monthFirst written 27 Oct. When what people say changes, it shows here.
Tone· 55 posts from 48 people· written 27 Oct
How do people feel about Tavily?
Most of the 48 people here report Tavily doing the search job inside their own agents and apps (31 people), with 8 framing it as the invisible plumbing of agent search and 4 unimpressed.
Builders describe Tavily as the search step they no longer think about: an AI research agent built with Python and Groq, a medical RAG assistant locked to a domain allowlist, a data-migration tool that pulls outside references for a human to check. They name the same practical properties — structured results to drop straight into a prompt, framework and MCP integrations, and a Tavily MCP plus Crawl4AI pairing one calls "the combo I reach for first".
The infrastructure framing draws the reach: a Tech Week recap says most end users may never know the name while agent workflows need search underneath, and one poster tracks named customers such as CAVA's employee assistant, calling Tavily a "hidden gem".
The disappointed are few: a post calling the returned results dire next to what Google holds, a switch to Octen, and one calling it expensive against Exa. 2 people say it looks promising and intend to try it, without reporting results yet.
- Works well in my stack34
- Essential hidden infrastructure10
- Unimpressed4
- Free tier is enough4
- Promising, not yet tried3
See 5 of 55 postsHide the posts
BTW @cava is a Tavily customer that neither $NBIS nor Tavily appears to have announced. The restaurant chain has 476 locations and $365m in quarterly revenue. Tavily powers web research inside CAVA’s production employee assistant, Ask Astro. Employees can search internal company data and pull information from the web in the same conversation, with Tavily handling search and page extraction through MCP. link
Solid list — Tavily MCP + Crawl4AI is the combo I reach for first too; search quality and clean extraction cover 80% of researcher-agent needs. One thing I'd add from testing: the bottleneck is rarely the tools, it's goal drift — after reading 20 docs the agent forgets the original question. A notes/memory scratchpad between retrieval and synthesis helps more than a better search API. Docling is underrated here, by the way — PDF tables kill most RAG pipelines.
$NBIS' Tavily search agents search better. Probably worth keeping in mind if you’re building apps that need fast and precise search.
做 AI Agent 的人手一个的搜索 API，免费额度其实一直都在 Tavily 每月送 1,000 credits，注册不用绑卡，每月 1 号自动重置。普通搜索一次 1 credit，高级搜索 2 credit，自己跑小项目基本够用 它是 LangChain 这些框架默认接的搜索工具，返回的是清洗过的结构化数据而不是原始 HTML，给 Agent 联网查资料或者做 RAG 特别省事 免费层叫 Researcher，是每个月都给的循环额度，不是一次性的羊毛
5 of the 55 posts behind this answer, one per person. The other 50, and every new one as it’s posted, come with tracking. Get early access
Pain points· 16 posts from 15 people· written 27 Oct
What do people complain about with Tavily?
Price is the complaint most people in this section make — 8 of 15 say Tavily costs more than rivals per 1,000 searches — while 5 find the results weak or noisy and 2 hit API errors.
Rate cards are the argument: 9 posts price Tavily at $8 per 1,000 basic searches and up to $16 for advanced, against $0.75 for DeepSeek's new Search API, $1.00 for Parallel Turbo and $7.00 for Exa Auto, with Tavily Extract Advanced at $3.20 per 1,000 pages. One developer runs the arithmetic at a million searches — $8,000 against $750 — and another sorts search into three price bands, with Tavily in the middle of the three.
For solo builders the drain shows up as credits: advanced searches draw double credits, one says the free monthly allowance disappears fast, and another counts 276 requests burning 4,243 credits. On quality, 5 posts call results shallow or noisy — 82% fact recall against Firecrawl's 97% in one agent test, about half of a multi-asset image-and-video query irrelevant media, and a voice-agent builder saying web search "still needs a lot of improvement". 2 posts report API errors: one a status update mid-outage, one after hunting for the exact word a finished job returns.
- Costs more than rivals9
- Weak or noisy results5
- API errors and rough edges2
See 5 of 16 postsHide the posts
DeepSeek just made web search for AI agents 10x cheaper. And if you're building RAG systems or AI agents, you should probably pay attention. DeepSeek recently launched its own Search API. The price? $0.75 per 1,000 searches. For comparison: → DeepSeek: $0.75 → Parallel Turbo: $1.00 → Exa Auto: $7.00 → Tavily Basic: $8.00 Same 1,000 search requests. Very different bills. At 1 million searches: DeepSeek: $750 Tavily: $8,000 That's $7,250 in potential savings before accounting for differences in usage and features. But cheaper doesn't automatically mean better. Here's how I'd think about it: • DeepSeek: High-volume, cost-sensitive agents. Check the data policy first. • Parallel: Low-cost search with different quality tiers. • Exa: When retrieval quality and content features justify spending more. • Tavily: Mature integrations and straightforward agent workflows. And here's the part most developers miss. Your search API isn't your only cost. Every search result becomes context your LLM has to read. More results → more input tokens → higher inference costs. A $0.75 search API can still become expensive if your agent keeps stuffing thousands of unnecessary tokens into every prompt. So don't just benchmark search price. Benchmark retrieval quality, latency, tokens returned, and cost per correctly answered question. The cheapest search isn't always the cheapest agent. Which search API are you using right now? #AI
Your AI agent pays for two things: the model's tokens and the web searches it runs. Token prices fell about 10x a year. Search prices didn't follow. They split into three tiers: $0.75 to $1 at DeepSeek, Perplexity and Parallel, $4 to $8 at Exa, Brave and Tavily, and $10 to $14 for the search bundled with OpenAI, Anthropic and Gemini. On a cheap model, that top tier costs your agent more than the tokens. One bundled search costs a penny, the same as reading 100,000 tokens. Tokens are a hardware business. Search is an access business. link
Tavily 这个我用过 1000 次跑小 agent 确实够 不过高级搜索 2 credit 一不留神就见底了 平时都是先用普通的
BENCHMARK: Firecrawl vs Tavily for an AI agent. A test across 10 companies and 20 pages showed Firecrawl hitting 97% fact recall versus Tavily's 82%. Firecrawl was ~5.6x slower and ~5x the credit cost, but retrieval breadth directly impacts agent grounding.
I’ve been using @tavilyai to collect public articles for @Bharat_Icons with the 4k/mo student credits, and it works great. However, querying multi-asset searches like "C.V. Raman rare images, YouTube videos, articles" pulls too much noise—about 50% irrelevant media that’s painful to clean up automatically. @mynameisyahia How does Context .dev optimize search queries to filter irrelevance compared to Tavily? Does it handle structured multi-modal retrieval better? i am very interested here as i took "MULTIMODAL INFORMATION RETRIEV" in my 3rd sem
5 of the 16 posts behind this answer, one per person. The other 11, and every new one as it’s posted, come with tracking. Get early access
Alternatives· 32 posts from 28 people· written 27 Oct
What are the alternatives to Tavily?
The rival people weigh Tavily against most is Exa (8 of 28 people), largely on price and benchmark scores; DeepSeek's new Search API is next (4), on a rate quoted about ten times cheaper.
On price, one chart puts Exa Auto at $7.00 per 1,000 searches against Tavily Basic at $8.00, and one developer calls Exa "cheaper and best" while calling Tavily expensive. On quality the argument splits: a builder's own 100-check benchmark, run with a team of AI bots, has Tavily passing 69 checks to Exa's 40, revised to 75 and 57 under stricter URL rules; another recall test scores Tavily 60.9% against Exa's 79.3%; a developer who reproduced the first run says Tavily stays but Exa and Firecrawl answer 95% of the queries.
DeepSeek's Search API at $0.75 per 1,000 is the second comparison — one says DeepSeek's quality is worse than Tavily's, and a solo developer stays on Tavily for the free monthly credits and easy MCP setup. Firecrawl divides on billing, per request versus per page, and one test puts it at 97% fact recall to Tavily's 82% but 5.6x slower. Brave and the search bundled into OpenAI, Anthropic and Gemini draw 3 people each; Context.dev 2, then single mentions.
- Exa12
- DeepSeek Search4
- Firecrawl3
- Brave3
- Built-in model search3
- Context.dev2
See 5 of 32 postsHide the posts
AI替你搜索，钱都给了谁？5分钟讲透 同样两年下载涨了六十倍：自建网页库的 Exa 估值 22 亿美元；转手的 Tavily 被买走，SEC 文件写着成交 1.89 亿，不是报道的「最多 4 亿」；最好的默认位置在模型公司手里，Claude 的编程工具文档写着搜索后台「不能换」。 口诀：看一家 AI 搜索公司，问三句——有没有自己的网页库？是谁的默认？网站认不认它的爬虫？
We built one: link Oct 23 run: Tavily passed 69 of 100 of our search checks, Exa 40 of 100. Firecrawl wasn't in this run, so no number yet for your agents' favorite. One run, early signal, not a verdict. Made by me and my team of AI bots.
DeepSeek just made web search for AI agents 10x cheaper. And if you're building RAG systems or AI agents, you should probably pay attention. DeepSeek recently launched its own Search API. The price? $0.75 per 1,000 searches. For comparison: → DeepSeek: $0.75 → Parallel Turbo: $1.00 → Exa Auto: $7.00 → Tavily Basic: $8.00 Same 1,000 search requests. Very different bills. At 1 million searches: DeepSeek: $750 Tavily: $8,000 That's $7,250 in potential savings before accounting for differences in usage and features. But cheaper doesn't automatically mean better. Here's how I'd think about it: • DeepSeek: High-volume, cost-sensitive agents. Check the data policy first. • Parallel: Low-cost search with different quality tiers. • Exa: When retrieval quality and content features justify spending more. • Tavily: Mature integrations and straightforward agent workflows. And here's the part most developers miss. Your search API isn't your only cost. Every search result becomes context your LLM has to read. More results → more input tokens → higher inference costs. A $0.75 search API can still become expensive if your agent keeps stuffing thousands of unnecessary tokens into every prompt. So don't just benchmark search price. Benchmark retrieval quality, latency, tokens returned, and cost per correctly answered question. The cheapest search isn't always the cheapest agent. Which search API are you using right now? #AI
DeepSeek 开放了搜索 API，1000 次搜索只要 5 元。 价格确实很有竞争力。Artificial Analysis 的评测中，DeepSeek 得分 69，Tavily Basic 得分 66。 但作为个人开发者，我可能还是会继续用 Tavily。 原因很简单：Tavily 每月有 1000 个免费 credits，我目前基本够用，而且 MCP 接入也方便。 对于个人开发者，API 单价再便宜，也未必比免费更有吸引力。 DeepSeek 的低价策略可能更适合高调用量的企业场景。 大家的 Agent 都在用什么搜索服务？付费么？ 相关链接： DeepSeek Search API： link Artificial Analysis 搜索评测： link Tavily 定价： link
在选择 Agent 使用的搜索 API 时，许多人会先看价格，例如 Firecrawl 的每个 credit 看起来比 Tavily 便宜近十倍。 我一开始也是这样想的，后来发现同一个查询，两家的账单可以差六十多倍。但在一周真实负载的平均之后，便宜的反而变成了另一家。 根本原因是两家的计价模式天生不同：Tavily 按你发的请求计费，查到啥都不加钱；Firecrawl 按返回内容计费，普通网页每页一个 credit，PDF 按页数，X/Twitter 每条用 Grok 固定 30 个点。 所以选型不应该只看价格，而应该看你自己的负载里有什么内容。HTML 为主的日常检索，Firecrawl 换算成钱更省；长 PDF、社交动态占比一上来，Tavily 反超。文章用一周真实用量做了双向回放，把这条线画了出来。 link
5 of the 32 posts behind this answer, one per person. The other 27, and every new one as it’s posted, come with tracking. Get early access
Questions· 7 posts from 6 people· written 27 Oct
What do people ask about Tavily?
The questions are few and practical (6 people): 2 ask whether Tavily's terms cover storing snippets such as Reddit content for AI output, 2 ask what search service other developers use for agents and whether they pay, and the rest ask for comparisons.
The licensing question runs in two directions: one person asks a builder using Tavily to discover Reddit content whether they needed separate approval from Reddit to store snippets and reuse them in AI drafts, and the builder answers that they are still looking into the licensing implications. Two more ask what other developers use — one directly naming Tavily, Firecrawl, Crawl4AI and SearXNG as the options for giving local LLMs web access, and a developer weighing DeepSeek's cheaper API who ends by asking what everyone's agent uses and whether they pay for it.
The remaining questions are comparison shopping: whether anyone has replaced Tavily with Parallel and how it went on cost and quality, a request to compare Tavily and Exa side by side, and a search for the best open-source alternative covering web search, crawling and content extraction.
- Are stored snippets covered?2
- What search service do you use?2
- Has anyone replaced it with Parallel?1
- How does it compare with Exa?1
- What's the open-source alternative?1
See 5 of 7 postsHide the posts
congrats on launching loudship mate! curious how you handle reddit discovery through tavily did you need separate approval from reddit to store snippets and use them for ai drafts or does tavily agreement cover that?
Thanks mate! Really appreciate it! Yes, I'm using Tavily for discovery. I haven't requested separate approval from Reddit, and I'm still looking into the exact licensing implications around storing snippets and using them for AI-generated drafts. Definitely something worth clarifying as Loudship grows!
Genuine question 👀 If ChatGPT and Perplexity can search the web for you… what do developers use to give their local LLMs web access? Tavily, Firecrawl, Crawl4AI & SearXNG ? What’s your go-to? 👇
DeepSeek 开放了搜索 API，1000 次搜索只要 5 元。 价格确实很有竞争力。Artificial Analysis 的评测中，DeepSeek 得分 69，Tavily Basic 得分 66。 但作为个人开发者，我可能还是会继续用 Tavily。 原因很简单：Tavily 每月有 1000 个免费 credits，我目前基本够用，而且 MCP 接入也方便。 对于个人开发者，API 单价再便宜，也未必比免费更有吸引力。 DeepSeek 的低价策略可能更适合高调用量的企业场景。 大家的 Agent 都在用什么搜索服务？付费么？ 相关链接： DeepSeek Search API： link Artificial Analysis 搜索评测： link Tavily 定价： link
Anyone successfully replaced Tavily with Parallel for web search in AI agents and workflows? How's the experience been so far, particularly cost-wise and quality-wise?
5 of the 7 posts behind this answer, one per person. The other 2, and every new one as it’s posted, come with tracking. Get early access
In their words
How do people talk about Tavily?
The phrases people use most, counted once per person.
- 1“web search”7
- 2“search api”6
- 3“ai agents”4
- 4“research agent”4
- 5“agentic search”2
- 6“search quality”2
- 7“deep research”2
- 8“separate approval”2
Posts are shown the way X shows them. One of them yours? Here’s why it’s shown, and how to take it off.
Also in Web data for AI
- Apify—Reception is warm and practical: as many people vouch for Apify outright as describe a real job it delivered.247 posts / 7d
- Firecrawl—The biggest thread is the output: people praise clean markdown and structured JSON an agent can read, while the free US government records index draws the questions and wishes.232 posts / 7d
- Browserbase−48%Reception is warm: plain praise leads at 11 people and real work follows at 8, with login handling the feature people name.201 posts / 7d
- Exa—Praise is the largest theme on the page — 20 people, many of them running agents on it — and next comes excitement over the pairing with Pyth's market data (12 people).139 posts / 7d
Get every post about Tavily, filtered and labelled
This page is the summary. Tracking gives you the stream behind it for any brand, keyword or @mention: each new pain point, request and question as it’s posted, with the spam and promotions already taken out.