Skip to main content

Perplexity connector

Back to the connector overview, authentication methods, and actions.
​
17 Sept 2026
description-changesClarified the chat completion, web search, agent, and async tool descriptions so integrations pick the right action and build valid arguments. Web search returns links only while chat completion returns written answers (including SEC and academic search modes); the agent action is for multi-step orchestration and for specific openai/anthropic models or fallback chains; async is only for sonar-deep-research. Chat completion’s model is limited to the four Sonar values and its messages must be a real array, and heavy deep-research work is routed straight to the async or agent background paths instead of the sync call that can hit the gateway timeout. Agent responses that select an Anthropic model now default max_output_tokens automatically when it is not supplied, since Anthropic models reject the request without it.
perplexity_create_agent_response, perplexity_create_async_chat_completion, perplexity_create_chat_completion, perplexity_retrieve_agent_response, perplexity_search_web
​
16 Sept 2026
data-syncAdded scheduled data sync to the models and async chat completions list actions so these collections can be kept in sync and read back without calling Perplexity on every request. Async listing now accepts optional limit and next_token cursor params; both actions refresh in full each run, since neither endpoint offers an updated-since filter.
perplexity_list_async_chat_completions, perplexity_list_models
​
16 Jul 2026
bug-fix guide-fixFixed connection validation - switched the credential test action to list_async_chat_completions because the previous list_models endpoint is unauthenticated, so invalid API keys were accepted. Corrected the connection guide (API key field description, Buy more credits prerequisite, copy-icon label).