Perplexity with CAMEL-AI

Use the Perplexity Search API as a tool inside CAMEL-AI multi-agent systems.

Overview

CAMEL-AI’s SearchToolkit ships a search_perplexity method that wraps the Perplexity Search API as a native CAMEL tool. The tool returns ranked web results with a title, URL, and snippet, plus publication and update dates when available, so the calling ChatAgent can ground its responses on fresh, citable sources.

CAMEL-AI is an open-source framework for building communicative multi-agent systems. Learn more at camel-ai.org.

Installation

pip install "camel-ai==0.2.91a7"

Pin the prerelease. The latest stable release, 0.2.90, does not include SearchToolkit.search_perplexity. 0.2.91a7 is the latest prerelease and it includes CAMEL’s compatibility bounds for mcp and openai, so it installs without extra pins.

API Key Setup

Set your Perplexity API key as an environment variable. ChatAgent also requires an API key for its model provider. CAMEL uses OpenAI by default, so the example below sets OPENAI_API_KEY. If you configure another model provider, set that provider’s API key instead.

export PERPLEXITY_API_KEY="your_api_key_here"
export OPENAI_API_KEY="your_openai_api_key_here"

Get API Key

Generate your API key from the Perplexity dashboard.

Quick Start

Register search_perplexity as a tool on a ChatAgent:

from camel.agents import ChatAgent
from camel.toolkits import FunctionTool, SearchToolkit

agent = ChatAgent(
    system_message=(
        "You are a helpful assistant that can use the Perplexity "
        "search engine to answer questions."
    ),
    tools=[FunctionTool(SearchToolkit().search_perplexity)],
)

response = agent.step(
    input_message="What are the latest developments in multi-agent AI systems?",
    response_format=None,
)
print(response.msgs[0].content)

The tool calls POST https://api.perplexity.ai/search under the hood and returns the raw API response so the agent can pick which results to cite.

Configuration

search_perplexity provides first-class arguments for query, max_results, max_tokens, max_tokens_per_page, country, search_recency_filter, search_domain_filter, and search_language_filter. Pass other Search API fields through extra_params.

from camel.toolkits import SearchToolkit

toolkit = SearchToolkit()

results = toolkit.search_perplexity(
    query="AI safety research",
    max_results=5,
    max_tokens_per_page=512,
    country="US",
    search_recency_filter="week",                         # hour | day | week | month | year
    search_domain_filter=["arxiv.org", "nature.com"],
    search_language_filter=["en"],
)

Domain filters can use either allowlist entries such as "arxiv.org" or denylist entries such as "-medium.com", but you cannot mix allowlist and denylist modes in one request.

query can also be a list of up to five queries for batched search:

from camel.toolkits import SearchToolkit

toolkit = SearchToolkit()

results = toolkit.search_perplexity(
    query=["NVIDIA earnings", "AMD earnings"],
    max_results=3,
    extra_params={
        "search_after_date_filter": "1/1/2026",
        "search_before_date_filter": "5/1/2026",
    },
)

Response Shape

search_perplexity returns the raw Perplexity Search API response as a Dict[str, Any]. On success, id and results are always present. server_time is optional: the response can omit it or set it to None. Each result requires title, url, and snippet; date and last_updated are optional and can be None.

{
    "id": "...",
    "server_time": "...",  # optional; may be omitted or None
    "results": [
        {
            "title": "...",
            "url": "...",
            "snippet": "...",
            "date": "...",          # optional; may be None
            "last_updated": "...",  # optional; may be None
        },
        # ...
    ],
}

If PERPLEXITY_API_KEY is missing, CAMEL raises ValueError before making a request. Non-200 responses, request exceptions, and response-decoding errors return a dictionary with an error key.

Links & Resources

CAMEL-AI Docs

Build multi-agent systems with CAMEL-AI.

Tool Source

SearchToolkit.search_perplexity implementation.

Search API Quickstart

Learn more about the underlying Perplexity Search API.

Example Script

End-to-end CAMEL example using Perplexity search.

Support

Need help with the integration?

• Browse the CAMEL-AI documentation
• Review our FAQ