Perplexity · Technologies
Sonar API
Web grounded answers with citations, inside your product.
The Sonar API puts Perplexity’s search grounded models behind your own software: every response is answered from the live web with the citations built in. It is OpenAI compatible, so it drops into existing codebases, and it spans fast lookups to deep autonomous research. We build cited research into products and workflows.
GROUNDED WORKLOADS · EXAMPLE PLATFORM
WORKLOAD
TIER
GROUNDING
STATUS
Fact-Lookup
Fast
Cited
LIVE
Account-Research
Pro
Cited
LIVE
Market-Reports
Deep research
Cited
LIVE
Cost-Controls
Search fees
Tuned
REVIEW
ungrounded-answers
No sources
None
RISK
Sample platform
live · review · risk
In plain terms
Language models know the past. Sonar reads the present.
One API call that searches, answers and cites, instead of a pipeline you maintain. Here is what changes.
Without it
- A search API, an LLM and citation logic stitched by hand
- Products answering from training data that has aged
- Users asked to trust answers with no sources shown
- Freshness problems discovered by your customers
With it
- Search, answer and citations in a single call
- Responses grounded in the live web, not a cutoff
- Sources attached to every answer your product gives
- A research layer that stays current by design
What CG TECH can do with the Sonar API
The work, broken into the parts that matter.
Citations are the product
Every Sonar response is answered from live web sources with the references included, which is exactly what compliance heavy and trust sensitive products need from AI.
every answer arrives with sources
Fast lookups to deep research
The Sonar family spans quick factual answers, multi source synthesis, step by step reasoning and autonomous deep research that reads widely and reports. We route each workload to the tier that earns its cost.
the right depth per question
OpenAI compatible, deliberately
Sonar speaks the API format your codebase already knows, so adding cited research is a model string change, not a rebuild. It pairs naturally with Claude or GPT handling the reasoning.
cited research without a rebuild
Search fees, managed
Sonar bills per token plus per request search fees, and deep research decides its own search count, so costs need designing, not discovering. We build the budgets, tiers and monitoring in from day one.
a research bill you saw coming
How an engagement runs
From stale answers to a grounded product, step by step.
01
Discover
We map where your product or workflow needs current, sourced answers.
02
Design
Tier routing, cost controls and citation handling designed up front.
03
Build
The workload shipped with monitoring and budgets in place.
04
Handover
Runbooks, dashboards and a tuning rhythm your team owns.
Questions we hear a lot
Common questions about the Sonar API
What is the Sonar API?
The Sonar API puts Perplexity search grounded models behind your own software. Every response is answered from the live web with the citations built in. It is OpenAI compatible, so it drops into an existing codebase, and it spans fast lookups through to deep autonomous research.
What makes Sonar different from the OpenAI or Claude APIs?
Grounding. Sonar answers from a live web search with citations included in the response. General models reason and draft brilliantly but answer from training data. Many production systems use both: Sonar to find and cite, another model to reason.
Is it hard to integrate?
No. The API is OpenAI compatible, so most codebases swap it in with configuration rather than rework. The real work is designing tiers, citations and cost controls properly, which is what we do.
What does it cost?
Per token plus per request search fees that vary with search depth. Deep research runs its own searches, so we model costs on your real query mix during discovery rather than quoting a brochure number.
Where does it fit best?
Anywhere the answer must be current and sourced: account and market research tools, due diligence workflows, monitoring, and any product where showing sources builds the trust you sell on.
What does OpenAI compatible actually mean?
It means Sonar uses the same request and response shape as the OpenAI API. In most codebases you change the address you call and the model name, and the rest of your code keeps working.
Do the citations come back with the answer?
Yes. Sources come back alongside the response, so your product can show them to the person reading it. That is the main reason to choose Sonar over a model with no web access: the answer can be checked.
What should we build with it first?
Something where the answer has to be current and checkable. A research assistant inside your product, a market or competitor watch, or a screen that summarises what changed this week with links to the sources.
Ready when you are
Building cited research into your product? Let us talk.
A discovery session maps your use cases, your risks and your quick wins. You keep the plan either way.
What to expect
- A consultant replies within 4 business hours
- Session booked to understand your requirements
- We will provide you with a fixed price quote