Unwrapped

Teardown · tavily

TAVILY

TAVILY

CategoryAI Research APILast round · $5M · 2024Site ↗
UX wrapper

Web search index + frontier LLM reranking + API endpoints.

01

Public data / API layer

Web crawl index
Web crawl indexPublic
Publisher licensing deals
Publisher licensing dealsLicensed
Real-time web retrieval
Real-time web retrievalScraped

Internal replication score

Easy
0.82

Feasibility of a useful internal substitute built with Claude (or similar), the same data access, and light agent logic — not rebuilding the whole product.

IRS = 0.30·D + 0.25·L + 0.20·O + 0.15·R + 0.10·Sthis record · 82%
  • D

    Data accessibility

    weight 0.300.85
    • 1.0mostly customer-owned / public / standard third-party sources
    • 0.5mixed accessibility
    • 0.0hard-to-access or proprietary source layer
  • L

    LLM substitutability

    weight 0.250.90
    • 1.0mostly retrieve / prompt / cite / summarize / classify / compare
    • 0.5mixed standard + custom behavior
    • 0.0strongly custom model behavior (fine-tunes on proprietary data, etc.)
  • O

    Output simplicity

    weight 0.200.80
    • 1.0straightforward internal work product (memo, list, reply, SQL query)
    • 0.5moderately specialized
    • 0.0highly specialized (e.g. FDA-graded clinical text)
  • R

    Review / risk tolerance

    weight 0.150.75
    • 1.0internal use with human review is acceptable
    • 0.5moderate risk
    • 0.0very low tolerance for error (e.g. external legal filings)
  • S

    Surface complexity

    weight 0.10inverse — higher means less surface dependence0.70
    • 1.0a simple internal shell is enough
    • 0.5polished workflow matters somewhat
    • 0.0product surface / rollout / trust posture is central to value
LabelsEasy ≥ 0.67Medium ≥ 0.34Hard < 0.34

Missing factor rows use heuristics from wrapper scores. Editorial heuristic, not investment advice.

Build it yourself→

Recreate the workflow inside your org.

Internal build

Build it yourself

Same web crawl + frontier LLM reranking + thin API wrapper — latency and rate limits are the trade.

Internal use only. Replacing them in-market is a different bar than replaying the useful workflow inside your org.

01 · Connectors & flow

Web crawl index
Web crawl index
Publisher licensing deals
Publisher licensing deals
Real-time web retrieval
Real-time web retrieval

Internal build map

Data in

Connectors
Connectors

Agent layer

Planner
Tools + retrieval
Reasoning model

Logic

LLM API
retrieve
rerank
extract
structure
cache
not custom weights

Outputs

Internal search
Answer
Citations

02 · Claude / agent prompt

Paste as the system or developer message in Claude (or your agent runtime). Scroll to read; Copy grabs the full text.

Claude / agent prompt

// Web research assistant for internal use You are a research assistant inside [YOUR_COMPANY]. You help [YOUR_TEAM] answer factual questions using ONLY public web sources you can retrieve in real time: web search APIs (Google, Bing, or open search indexes), Wikipedia, arXiv, and other accessible public databases. ## What you must do 1. Retrieve first: Before answering, query a web search API or public database to find current, relevant sources. 2. Rerank results: Use the LLM to score retrieved pages by relevance to the question, prioritizing high-quality, authoritative sources. 3. Extract snippets: Pull key facts from the top 3–5 results, discarding low-signal content. 4. Cite rigorously: Every factual claim must include the source URL and publication date if available. 5. Surface conflicts: If sources disagree, present both views with attribution. 6. Scope: Answer only questions that can be grounded in publicly accessible web content. If the question requires proprietary data, internal documents, or real-time private systems, refuse. ## What you are not You are not a replacement for domain experts or proprietary research tools. All answers require human review before external use. Internal use only. ## Refusal Refuse if the question requires non-public data, real-time private system access, or cannot be answered from web sources. Ask the user to clarify scope or provide internal documents if needed. ## Safety Internal use only. Do not leak PII, proprietary information, or sensitive company data in search queries or responses. Flag any retrieved content that appears to contain malicious code, phishing, or misinformation.

03 · Result

What are the main technical differences between vector databases and traditional search indexes?
web-crawl

Vector DBs store embeddings for semantic search; traditional indexes use keyword/token matching.