Last updated: 11/17/2026Last verified: 2026-11-13
Octen is a real-time web search API platform for LLMs and AI agents: Web, Broad, Pulse, image, news, business and place search, URL extraction, embedding models and cited-answer and deep-research endpoints, with pay-as-you-go pricing and a $20/month Search Plan for unlimited Web Search.
What is Octen?
Octen (Octen Inc.) is a real-time web search API platform for LLMs, AI agents and apps. One API key covers live web search, specialized image, news, business and place search, URL extraction, text and vision-language embeddings, and higher-level endpoints for cited answers and deep research. Developers connect it through the REST API, a Python SDK, a CLI, an MCP server or Octen Skills for coding agents such as Claude Code and Cursor.
Key Features
- Web Search: ranked live-web results with highlights and optional full content.
- Broad Search: splits one question into parallel sub-queries and searches them in a single call.
- Pulse Search: the Octen-WebGR generative retrieval model generates up to 10 likely URLs for a query; the docs give server-side P50 latency of about 10ms in lite mode and 23ms in standard mode, and recommend Web Search for news or fresh pages.
- Image, News, Business and Place Search endpoints, with Video Search in early access.
- Extract: turns a URL into clean Markdown with intent-focused highlights and page classification.
- Text Embedding (0.6B, 4B and 8B) and VL Embedding APIs for retrieval across text, images, video and documents.
- Answer, Search Agent, Deep Research and Grounded Generation endpoints built on Octen search.
- Octen Skills, an MCP server, a CLI and a Python SDK for Claude Code, Cursor, VS Code and other MCP clients.
Best For
Pricing
Octen bills pay-as-you-go from a prepaid balance; new accounts get $5 in free balance and a free 10 QPS tier (20 QPS after adding credits). Per the pricing page and docs read 2026-11-13: Web Search and Broad Search cost $1 per 1,000 calls (shown as 80% off $5), Pulse Search $0.1 (lite) or $0.2 (standard) per 1,000 calls, Image Search $2, News Search $3 and Business or Place Search $5 per 1,000 calls; full content beyond the free results is $0.5 per 1,000 results, and Extract is $1 (standard) or $2.5 (advanced) per 1,000 successful URLs. The Search Plan makes Web Search unlimited for $20/month ($50/quarter or $180/year) with 1,000 Image Searches a month; Plus is $100/month with 5,000. Search Plans are for personal prepaid accounts, may queue at busy times, and do not cover other APIs or full content. QPS Plans with guaranteed throughput start at $2,099/month (Builder, shown 30% off $2,999).
Pros & Cons
Pros
- Search, extraction, embeddings and research endpoints from one provider and one API key.
- Low published per-call prices plus a flat-price unlimited Web Search plan.
- Several integration routes: REST API, Python SDK, CLI, MCP server and agent Skills.
- Docs are published in Markdown with an llms.txt index, which suits coding agents.
- Pulse Search offers a very low-latency option for navigational lookups.
Cons
- The $20 Search Plan covers Web Search only; requests may queue at busy times and guaranteed throughput needs an expensive QPS Plan.
- Pulse Search targets docs, official sites and popular destinations; for news or fresh pages the docs point to Web Search.
- The open Octen-WebGR weights on Hugging Face are CC BY-NC 4.0, so self-hosting them is non-commercial; commercial use goes through the hosted API.
- Benchmark and leaderboard figures on the site are cited by Octen itself; this overview did not create an account or call the API.
Alternatives
FAQ
Details
Platform
Features
- Web Search and Broad Search APIs return ranked live-web results with highlights and optional full content; Broad Search splits one question into parallel sub-queries in a single call
- Pulse Search uses the Octen-WebGR generative retrieval model to generate up to 10 likely URLs for a query; the docs give server-side P50 latency of about 10ms in lite mode and 23ms in standard mode, and suggest Web Search for news or fresh pages
- Extract turns a URL into clean Markdown with highlights and page classification; text and vision-language embedding APIs plus Answer, Search Agent, Deep Research and Grounded Generation endpoints are built on the same search
- Works from Claude Code, Cursor, VS Code and other MCP clients through Octen Skills, an MCP server and a CLI, with a Python SDK
Languages
Known limitations
- Billing is prepaid pay-as-you-go from an account balance (new accounts get $5 free balance and start on a free 10 QPS tier). Per the pricing page read 2026-11-13, Web and Broad Search are $1 per 1,000 calls (shown as 80% off $5), Pulse Search $0.1 (lite) or $0.2 (standard) per 1,000 calls, and full content beyond the free results costs $0.5 per 1,000 results
- The $20/month Search Plan (also $50/quarter or $180/year; Plus is $100/month) covers unlimited Web Search only, plus 1,000 Image Searches a month. It is for personal prepaid accounts, may queue at busy times, and other APIs and full content stay pay-as-you-go; guaranteed throughput needs a QPS Plan starting at $2,099/month (shown 30% off $2,999)
- The open Octen-WebGR models on Hugging Face (4B and 0.6B) are released under CC BY-NC 4.0, so self-hosting is non-commercial; the model card says commercial use goes through the hosted Pulse Search API. Pulse Search is aimed at docs, official sites and popular destinations rather than news or fresh pages
- Speed, cost and leaderboard figures on the site (for example the Artificial Analysis Search API Benchmark and RTEB/MMEB rankings) are reported by Octen; this listing did not create an account or call the API








