Perplexity connector for AI agents

Tools your research agent reaches for on Perplexity, scoped per user.

CALL ANY TOOL

17 tools for web search in Perplexity: search the web, get Sonar answers with citations, run sync and async completions, run and cancel Agent API responses, create embeddings, evaluate content, and list custom skills.

perplexity_answer_decisions

Answer decisions

Evaluate named yes/no, multiple-choice, and scoring questions against a piece of content. Returns an answers object keyed by question name (probability for noul, chosen option for choice, level for score), plus the model and token usage. Use this for classification, moderation, routing, and rubric scoring over text; use perplexity_chat_completion for open-ended answers.

Parameters

Name

Type

Required

Description

model

string

Required

Decision model: pplx-decider-v1.2 (larger) or pplx-decider-flash-v1.2 (faster, cheaper). Aliases pplx-decider-latest and pplx-decider-flash-latest also work.

questions

object

Required

Questions as a JSON object mapping a name you choose to a question. Each question has a type of noul (yes/no, with instructions), choice (with criteria mapping option names to descriptions), or score (with criteria listing level descriptions). Example: {"is_bug":{"type":"noul","instructions":"Is this a bug report?"},"area":{"type":"choice","criteria":{"billing":null,"settings":null}}}

state

string

Required

The content every question refers to: plain text, or a JSON object or array serialized as a string. Example: Customer wrote: the app crashes every time I open settings.

perplexity_cancel_agent_response

Cancel agent response

perplexity_chat_completion

Chat completion

Generate a web-grounded answer from a Sonar model for a conversation (non-streaming). Returns the assistant message in choices, plus citations, search_results, token usage, and optional images and related_questions. Use this for quick cited answers. Use perplexity_create_agent_response for multi-step agentic work, perplexity_create_async_chat_completion for slow sonar-deep-research jobs, and perplexity_search for raw results.

Parameters

Name

Type

Required

Description

messages

array

Required

Conversation so far as a JSON array of message objects, each with a role (system, user, assistant, or tool) and a content string. End with a user message. Example: [{"role":"user","content":"What changed in the EU AI Act this year?"}]

model

string

Required

Sonar model to use: sonar (fast, cheap), sonar-pro (deeper, more citations), sonar-reasoning-pro (step-by-step reasoning), or sonar-deep-research (long multi-step research, slow). One of: `sonar`, `sonar-pro`, `sonar-deep-research`, `sonar-reasoning-pro`.

disable_search

boolean

Optional

When true, turns web search off and the model answers from its training data only.

enable_search_classifier

boolean

Optional

When true, a classifier decides whether the query needs web search at all.

language_preference

string

Optional

ISO 639-1 code for the preferred response language. Example: en

last_updated_after_filter

string

Optional

Only use sources last updated after this date, MM/DD/YYYY. Example: 01/15/2026

last_updated_before_filter

string

Optional

Only use sources last updated before this date, MM/DD/YYYY. Example: 06/30/2026

max_tokens

integer

Optional

Maximum number of completion tokens to generate (up to 128000). Example: 1024

reasoning_effort

string

Optional

How much effort the model spends reasoning (minimal, low, medium, high); applies to reasoning-capable models. One of: `minimal`, `low`, `medium`, `high`.

response_format

object

Optional

Structured output control as a JSON object. Omit for plain text. For JSON output use {"type":"json_schema","json_schema":{"schema":{...}}}, where schema is a JSON Schema object.

return_images

boolean

Optional

When true, include image results in the response.

return_related_questions

boolean

Optional

When true, include suggested follow-up questions.

search_after_date_filter

string

Optional

Only use sources published after this date, MM/DD/YYYY. Example: 01/15/2026

search_before_date_filter

string

Optional

Only use sources published before this date, MM/DD/YYYY. Example: 06/30/2026

search_domain_filter

array

Optional

Limit search to these domains as a JSON array of domain strings. Example: ["wikipedia.org","github.com"]

search_language_filter

array

Optional

Filter results by language as a JSON array of ISO 639-1 codes. Example: ["en","fr"]

search_mode

string

Optional

Source of search results: web (default), academic for scholarly sources, or sec for SEC filings. One of: `web`, `academic`, `sec`.

search_recency_filter

string

Optional

Only use sources published within this time window. One of: `hour`, `day`, `week`, `month`, `year`.

stop

array

Optional

Stop sequences as a JSON array of strings; generation ends when one is produced. Example: ["END"]

temperature

number

Optional

Sampling randomness from 0 to 2; higher is more random. Example: 0.2

top_p

number

Optional

Nucleus sampling cutoff from 0 to 1. Example: 0.9

web_search_options

object

Optional

Web search tuning as a JSON object, for example {"search_context_size":"high"} (low, medium, or high amount of search context).

perplexity_create_agent_response

Create agent response

Run the Perplexity Agent API on a question or task, with optional web search tools, reasoning control, and structured output (non-streaming). Returns id, status, model, an output array of messages and tool calls, and token usage with cost. Use this for agentic, multi-step research; use perplexity_chat_completion for a quick cited Sonar answer. Set background to true and poll perplexity_get_agent_response for long runs. Provide either preset or model.

Parameters

Name

Type

Required

Description

input

string

Required

The question or task for the agent as plain text. Example: Compare the pricing of the three largest cloud vector databases.

background

boolean

Optional

When true the call returns immediately with status queued; poll perplexity_get_agent_response until the status is terminal. Use for long runs.

instructions

string

Optional

System instructions that set standing rules for the run. Example: Answer in three bullet points and cite sources.

language_preference

string

Optional

ISO 639-1 code for the response language. Example: en

max_output_tokens

integer

Optional

Maximum number of tokens to generate. Example: 2048

max_steps

integer

Optional

Maximum number of research loop steps (1-100); overrides the preset. Without a preset the default is 1. Example: 5

model

string

Optional

Model id in provider/model format, for example anthropic/claude-sonnet-5-5. Required when preset is not provided; use perplexity_list_models for valid ids.

models

array

Optional

Fallback chain of model ids in provider/model format, tried in order until one succeeds; takes precedence over model. Example: ["openai/gpt-5.6-terra","anthropic/claude-sonnet-5-5"]

preset

string

Optional

Preset name that bundles a model, system prompt, and search settings, such as fast, low, medium, high, or xhigh. Required when model is not provided.

previous_response_id

string

Optional

Id of an earlier completed response to continue the conversation from. Example: resp_abc123

reasoning

object

Optional

Reasoning control as a JSON object with an effort of minimal, low, medium, high, xhigh, or max. Example: {"effort":"high"}

response_format

object

Optional

Structured output as a JSON object: {"type":"json_schema","json_schema":{"name":"answer","schema":{...}}}, where name is 1-64 alphanumeric characters and schema is a JSON Schema object.

store

boolean

Optional

When false, the response is hidden from later retrieval; it can still be continued with previous_response_id. Default true.

temperature

number

Optional

Sampling temperature forwarded to the model. Example: 0.3

tools

array

Optional

Tools the agent may use, as a JSON array of tool objects, each with a type such as web_search, fetch_url, finance_search, people_search, image_search, or sandbox. Example: [{"type":"web_search"},{"type":"fetch_url"}]

top_p

number

Optional

Nucleus sampling cutoff forwarded to the model. Example: 0.9

perplexity_create_async_chat_completion

Create async chat completion

Submit a Sonar chat completion as a background job instead of waiting for the answer. Returns the job id, model, status, and timestamps; the finished answer appears in response once status is COMPLETED. Use this for long jobs such as sonar-deep-research. Poll with perplexity_get_async_chat_completion and browse jobs with perplexity_list_async_chat_completions.

Parameters

Name

Type

Required

Description

messages

array

Required

model

string

Required

disable_search

boolean

Optional

When true, turns web search off and the model answers from its training data only.

enable_search_classifier

boolean

Optional

When true, a classifier decides whether the query needs web search at all.

idempotency_key

string

Optional

Unique key so retries of the same submission do not create duplicate jobs. Example: report-2026-q3-eu-ai-act

language_preference

string

Optional

ISO 639-1 code for the preferred response language. Example: en

last_updated_after_filter

string

Optional

Only use sources last updated after this date, MM/DD/YYYY. Example: 01/15/2026

last_updated_before_filter

string

Optional

Only use sources last updated before this date, MM/DD/YYYY. Example: 06/30/2026

max_tokens

integer

Optional

Maximum number of completion tokens to generate (up to 128000). Example: 1024

reasoning_effort

string

Optional

How much effort the model spends reasoning (minimal, low, medium, high); applies to reasoning-capable models. One of: `minimal`, `low`, `medium`, `high`.

response_format

object

Optional

Structured output control as a JSON object. Omit for plain text. For JSON output use {"type":"json_schema","json_schema":{"schema":{...}}}, where schema is a JSON Schema object.

return_images

boolean

Optional

When true, include image results in the response.

return_related_questions

boolean

Optional

When true, include suggested follow-up questions.

search_after_date_filter

string

Optional

Only use sources published after this date, MM/DD/YYYY. Example: 01/15/2026

search_before_date_filter

string

Optional

Only use sources published before this date, MM/DD/YYYY. Example: 06/30/2026

search_domain_filter

array

Optional

Limit search to these domains as a JSON array of domain strings. Example: ["wikipedia.org","github.com"]

search_language_filter

array

Optional

Filter results by language as a JSON array of ISO 639-1 codes. Example: ["en","fr"]

search_mode

string

Optional

Source of search results: web (default), academic for scholarly sources, or sec for SEC filings. One of: `web`, `academic`, `sec`.

search_recency_filter

string

Optional

Only use sources published within this time window. One of: `hour`, `day`, `week`, `month`, `year`.

stop

array

Optional

Stop sequences as a JSON array of strings; generation ends when one is produced. Example: ["END"]

temperature

number

Optional

Sampling randomness from 0 to 2; higher is more random. Example: 0.2

top_p

number

Optional

Nucleus sampling cutoff from 0 to 1. Example: 0.9

web_search_options

object

Optional

Web search tuning as a JSON object, for example {"search_context_size":"high"} (low, medium, or high amount of search context).

perplexity_create_contextualized_embeddings

Create contextualized embeddings

perplexity_create_embeddings

Create embeddings

perplexity_get_agent_response

Get agent response

perplexity_get_async_chat_completion

Get async chat completion

perplexity_get_skill

Get skill

perplexity_get_skill_download_url

Get skill download url

perplexity_list_agent_response_files

List agent response files

perplexity_list_async_chat_completions

List async chat completions

perplexity_list_models

List models

perplexity_list_skill_revisions

List skill revisions

perplexity_list_skills

List skills

perplexity_search

Search the web and return ranked results with page titles, URLs, snippets, and dates. Returns a results array plus a request id; each result carries title, url, snippet, date, and last_updated. Use this for raw ranked results you will read yourself. Use perplexity_chat_completion or perplexity_create_agent_response for a synthesized answer with citations.

Parameters

Name

Type

Required

Description

query

string

Required

Search query text. For several related searches in one call, pass a JSON array of up to 5 query strings instead. Example: latest developments in solid-state batteries

country

string

Optional

Two-letter ISO 3166-1 alpha-2 country code to bias results toward. Example: US

last_updated_after_filter

string

Optional

Only return results last updated after this date, in MM/DD/YYYY format. Example: 01/15/2026

last_updated_before_filter

string

Optional

Only return results last updated before this date, in MM/DD/YYYY format. Example: 06/30/2026

max_results

integer

Optional

Maximum number of results to return (1-50). Values above 20 are only accepted with search_type "people". Default 10.

max_tokens

integer

Optional

Maximum total webpage content tokens to return across all results (1-1000000). Example: 25000

max_tokens_per_page

integer

Optional

Maximum webpage content tokens to extract from each result page (1-1000000). Example: 2048

search_after_date_filter

string

Optional

Only return results published after this date, in MM/DD/YYYY format. Example: 01/15/2026

search_before_date_filter

string

Optional

Only return results published before this date, in MM/DD/YYYY format. Example: 06/30/2026

search_domain_filter

array

Optional

Restrict results to these domains (max 20), as a JSON array of domain strings. Example: ["nature.com","arxiv.org"]

search_language_filter

array

Optional

Restrict results to these languages (max 20), as a JSON array of two-letter ISO 639-1 codes. Example: ["en","de"]

search_recency_filter

string

Optional

Only return results published within this time window. One of: `hour`, `day`, `week`, `month`, `year`.

search_type

string

Optional

Search mode: "web" for general web search (default), "fast" for lower-latency and lower-cost web search, "people" for people search. One of: `web`, `fast`, `people`.

Build your Agent

Same auth pattern across LangChain, OpenAI, Anthropic, and Google ADK.

import { ScalekitClient } from "@scalekit-sdk/node";
import { createReactAgent } from "@langchain/langgraph/prebuilt";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);

// Perplexity tools scoped to this user
const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["perplexity"], toolNames: [
    "perplexity_search",
    "perplexity_chat_completion",
    "perplexity_create_agent_response"] },
  pageSize: 100,
});

const agent = createReactAgent({ llm, tools });
await agent.invoke({ messages: [{ role: "user", content: "Find recent news about Acme Corp's funding with citations" }] });

import OpenAI from "openai";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);
const openai = new OpenAI();

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["perplexity"] }, pageSize: 100,
});

const res = await openai.chat.completions.create({
  model: "gpt-5",
  messages: [{ role: "user", content: "Find recent news about Acme Corp's funding with citations" }],
  tools,
});

// Execute the tool call with the user's vaulted Perplexity access token
await sk.tools.executeTool(res.choices[0].message.tool_calls[0], "user_123");

import Anthropic from "@anthropic-ai/sdk";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);
const anthropic = new Anthropic();

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["perplexity"] }, pageSize: 100,
});

const msg = await anthropic.messages.create({
  model: "claude-sonnet-5",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Find recent news about Acme Corp's funding with citations" }],
  tools,
});

// Tool call runs with the user's vaulted Perplexity access token
await sk.tools.executeTool(msg.content, "user_123");

import { Agent } from "@google/adk/agents";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["perplexity"] }, pageSize: 100,
});

const agent = new Agent({
  name: "research_agent",
  model: "gemini-2.5-pro",
  instruction: "Perplexity tools scoped to this user",
  tools,
});

await agent.run("Find recent news about Acme Corp's funding with citations");

Try these prompts

Copy any prompt into your agent. Each maps directly to a Perplexity tool. Click to copy, paste into your agent, done.

Search and answers

Copy the prompt

Copied

Search the web for agent authentication platforms launched this year.

Copy the prompt

Copied

Find recent news about Acme Corp's funding with citations.

Copy the prompt

Copied

Submit an async research question on EV battery suppliers.

Agent API

Copy the prompt

Copied

Run the Agent API on a comparison of 3 CRM vendors.

Copy the prompt

Copied

Check whether my agent response has finished.

Copy the prompt

Copied

List the files my last agent response generated.

Embeddings and evaluation

Copy the prompt

Copied

Create embeddings for these 100 support articles.

Copy the prompt

Copied

Does this blog post mention pricing? Answer yes or no.

Copy the prompt

Copied

List the custom skills in our Perplexity project.

SEE HOW AUTH WORKS

Each user connects their own Perplexity access token once, and Scalekit sends it with every call. Access tokens stay vaulted, every call is scope checked, and every action is logged.

Authorize

Your user connects

Perplexity

once. We tie it to their identity and the meetings they approved — no shared bot account, no org-wide access

Who:

user ‘A’

when:

Once per user

access:

Limited to user

Store

Their

Perplexity

token lives in a vault scoped to them. User A's meetings are never reachable by an agent acting for user B, even on the same connection

vault:

encrypted

scope:

per-user

tokens:

auto-refreshed

Resolve

When your agent calls a

Perplexity

tool, we fetch the right token server-side. It never touches your agent, never appears in the LLM context, never shows up in your logs

speed:

~40ms

check:

before every call

seen by:

nobody

Audit

Every

Perplexity

tool call is logged — who triggered it, which meeting was fetched, what came back. 90 days of history, tied to the user who authorized it

history:

90 days

export:

SIEM-ready

logged:

every call

Test other agents

See the same per-user auth pattern across other AI and search connectors.

GTM and RevOps Teams

Competitive intelligence briefing agent

Scans Gong calls for competitor mentions, matches each one to its Notion battlecard, and DMs every affected rep a single Slack digest per cycle. Every call runs as the PMM who owns the briefing, never a shared bot.

GTM and RevOps Teams

Sales call prep agent

Reads tomorrow's calls from Google Calendar, mines past Granola notes and Attio history for context, and delivers each rep a prep brief in Slack, scoped to the calls that rep actually owns.

Why Scalekit

Secure your agent's access. Connectors ship in minutes

01.

Shared tokens break per-user analytics

A shared Perplexity access token looks fine in a demo. In production every search and answer looks like one service account, and you cannot tell which user triggered it. Scalekit resolves the credential of the actual user who triggered the agent, never a shared bot.

// shared key
audit → bot_service_account

// scalekit
audit → user_abc ✓

02.

Authentication is not authorization

03.

Multi-tenancy is architectural

04.

Perplexity today. Ten connectors tomorrow.

“Our agents act across Salesforce, Gong, Google Drive, and more, on behalf of every customer. Scalekit behind the scenes meant we can keep adding tools without ever rebuilding how credentials or tool calling work.”

Venu Madhav Kattagoni

Head of Engineering / Von

FAQs

Frequently Asked Questions

Does the agent access Perplexity as the user or through a shared access token?

As the user. Each user connects their own Perplexity access token once, and Scalekit sends it with every call. Audit logs attribute every search and agent run to that user, not a shared service account.

Where is the Perplexity access token stored?

In Scalekit's managed AES-256 token vault, namespaced per tenant. To rotate it, the user reconnects with a new access token. Revocation is a single dashboard action. Credentials never appear in prompts, logs, or LLM context.

Can I limit what the agent does in Perplexity?

Yes. Filter by tool name in listScopedTools to expose only what you want, for example perplexity_search and perplexity_chat_completion without the Agent API tools. No Perplexity tool deletes data. Scalekit also enforces scope checks before every API call.

What happens when a user revokes Perplexity access?

The connection is invalidated on the next tool call. Subsequent requests for that user fail closed with a clear error. Other users in the tenant remain unaffected. The event is logged for audit.

Does the agent get sources with each answer?

Yes. perplexity_chat_completion returns the Sonar answer with citations and search results, and perplexity_search returns ranked results with page titles, URLs, snippets, and dates.