Perplexity Agent API is one of 17 APIs that Perplexity publishes on the APIs.io network, described by a machine-readable OpenAPI specification and an AsyncAPI event-driven specification.
Tagged areas include Agents. The published artifact set on APIs.io includes an OpenAPI specification, API documentation, an API reference, authentication docs, pricing, and an AsyncAPI specification.
This API exposes
5 operations
across 5 paths,
and defines 99 schemas.
It is described by OpenAPI 3.2.0, at version 1.0.0.
Requests are made against a single base URL, https://api.perplexity.ai.
The identity and technical contract details declared by the specification.
UrlContent
object
Content fetched from a URL
3 properties
3 required
ConnectorTool
object
A saved managed or custom connector scoped to the authenticated Project.
6 properties
3 required
InputContentPart
object
6 properties
1 required
FetchUrlTool
object
2 properties
1 required
SearchQueriesEvent
object
Search queries event (type: "response.reasoning.searchqueries"). Contains search queries being executed.
4 properties
3 required
SearchSource
string
Source of search results
CustomSkill
object
3 properties
2 required
Currency
string
Currency code for cost values
McpApprovalRequestInput
object
A prior mcpapprovalrequest output item resent as manual history. Resend thoughtsignature unchanged when it is present.
7 properties
5 required
SearchResultsOutputItem
object
3 properties
2 required
McpTool
object
Connects a user-supplied remote MCP server. Agent API discovers the server's tools when the request starts and calls them like native tools. deferloading: true…
8 properties
3 required
TextDoneEvent
object
Text done event (type: "response.outputtext.done"). Contains the final text content.
6 properties
6 required
SearchResult
object
A single search result used in LLM responses
7 properties
4 required
ImageSearchFilters
object
3 properties
InlineSkill
object
4 properties
4 required
ApprovalFilters
object
Per-tool approval filters. An absent or null member matches no tool. Matching always takes precedence over matching never, and an unmatched call requires appro…
2 properties
FinanceResult
object
A single structured finance result returned by the financesearch tool.
4 properties
2 required
SandboxResultsOutputItem
object
Result of a sandbox tool invocation. Contains the executed code and its output.
7 properties
2 required
ImageSearchTool
object
3 properties
1 required
OutputItemDoneEvent
object
Output item done event (type: "response.outputitem.done"). Emitted when an output item (message or tool call) completes.
4 properties
4 required
ResponseCompletedEvent
object
Response event Contains the full or partial response object.
3 properties
2 required
FunctionCallOutputImagePart
object
3 properties
2 required
ResponseCreatedEvent
object
Response created event (type: "response.created"). Contains the initial response object.
3 properties
2 required
Annotation
object
Text annotation (URL citation)
5 properties
FunctionCallOutputInput
object
7 properties
3 required
McpListToolsOutputItem
object
Tools discovered on one external MCP server when the request starts. Matches OpenAI's mcplisttools item.
6 properties
4 required
McpCallOutputItem
object
One tool call executed against an external MCP server, modeled on OpenAI's mcpcall item.
9 properties
5 required
SearchDomainFilter
object
1 property
ResponsesCost
object
Cost breakdown for a Responses API request
8 properties
4 required
FetchUrlResultsOutputItem
object
2 properties
2 required
RequireApproval
Approval policy for an MCP server or connector. A call matching always requires approval; otherwise a call matching never runs automatically; otherwise approva…
ApprovalFilter
object
Matches a tool when every present field matches. An empty filter matches every tool.
2 properties
PeopleSearchTool
object
People search tool configuration for the Agent API
3 properties
1 required
ToolUserLocation
object
User's geographic location for search personalization
5 properties
McpApprovalResponseInput
object
The application's decision for one pending mcpapprovalrequest. Every pending request in the selected history needs exactly one decision before any approved cal…
5 properties
3 required
JSONSchemaFormat
object
Defines a JSON schema for structured output validation
4 properties
2 required
ContentPart
object
3 properties
2 required
NamespaceToolDef
object
One function definition discovered by hosted external-tool search.
4 properties
2 required
ResponsesObjectType
string
Object type in API responses
FunctionCallOutputTextPart
object
2 properties
2 required
ResponsesRequest
object
20 properties
1 required
ResponseFormat
object
Specifies the desired output format for the model response
2 properties
1 required
EventType
string
SSE event type discriminator
InputTokensDetails
object
3 properties
1 required
ToolCallTokenUsage
object
5 properties
4 required
ResponseFileList
object
2 properties
2 required
McpCallInput
object
A prior mcpcall output item resent as manual history. It records history and does not trigger execution.
9 properties
5 required
Date
string
Input: MM/DD/YYYY, Output: YYYY-MM-DD
SearchResultsEvent
object
Search results event (type: "response.reasoning.searchresults"). Contains search results returned.
5 properties
3 required
ImageResult
object
A single image search result
5 properties
4 required
WebSearchTool
object
Web search tool configuration for the Responses API
7 properties
1 required
TextDeltaEvent
object
Text delta event (type: "response.outputtext.delta"). Contains incremental text content.
6 properties
6 required
FunctionCallOutputItem
object
7 properties
6 required
NamespaceTool
object
A response-only namespace containing complete external tool definitions.
4 properties
4 required
BuiltinSkill
object
2 properties
2 required
FetchUrlResultsEvent
object
URL fetch results event (type: "response.reasoning.fetchurlresults"). Contains fetched URL contents.
4 properties
3 required
McpListToolsInput
object
A prior mcplisttools output item resent as history. The API accepts and drops it because tool discovery runs again.
5 properties
1 required
FinanceSearchTool
object
Finance search tool configuration for the Agent API
1 property
1 required
InputContent
Message content - either a string or array of content parts
ResponseInProgressEvent
object
Response in progress event (type: "response.inprogress"). Emitted when response processing has started.
3 properties
2 required
RoleType
string
Role in a message
InputMessage
object
3 properties
3 required
OutputItemAddedEvent
object
Output item added event (type: "response.outputitem.added"). Emitted when a new output item (message or tool call) starts.
4 properties
4 required
DateFilters
object
5 properties
ResponseFailedEvent
object
Response failed event (type: "response.failed"). Contains error details when streaming fails.
3 properties
3 required
ProfileReference
object
3 properties
2 required
FunctionTool
object
5 properties
2 required
ResponsesUsage
object
Token usage and cost information for a Responses API request
6 properties
3 required
ErrorInfo
object
Error information returned when a request fails
3 properties
1 required
Input
Input content - either a string or array of input items
FunctionCallInput
object
5 properties
4 required
ReasoningStartedEvent
object
Reasoning started event (type: "response.reasoning.started"). Signals the model has started reasoning/searching.
3 properties
2 required
ToolSearchOutputItem
object
Complete public definitions matched by one hosted external-tool search.
7 properties
6 required
ReasoningConfig
object
1 property
Status
string
Status of a response or output item
ContentPartType
string
Type of a content part
ResponseStreamEvent
SSE stream event. Discriminate by the type field: - response.created: Initial response object - response.inprogress: Response processing started - response.com…
FinanceResultsOutputItem
object
Intermediate output item emitted when the financesearch tool runs. One item is emitted per tool invocation; the requested categories and tickers are echoed at…
4 properties
2 required
PeopleSearchResultsOutputItem
object
Intermediate output item emitted when the peoplesearch tool runs. Mirrors the shape of searchresults: the agent's generated queries plus a list of per-person r…
3 properties
2 required
ReasoningStoppedEvent
object
Reasoning stopped event (type: "response.reasoning.stopped"). Signals the model has finished reasoning/searching.
3 properties
2 required
SandboxTool
object
Sandbox tool configuration for the Responses API. Executes code in an isolated container during an Agent API request.
1 property
1 required
ResponseFile
object
One file shared from a response's sandbox via sharefile.
5 properties
5 required
McpApprovalRequestOutputItem
object
One MCP or connector tool call waiting for the application to approve or deny it. The call has not run. Answer with an mcpapprovalresponse input item in the ne…
7 properties
5 required
MessageOutputItem
object
5 properties
5 required
OutputTokensDetails
object
1 property
1 required
FetchUrlQueriesEvent
object
URL fetch queries event (type: "response.reasoning.fetchurlqueries"). Contains URLs being fetched.
4 properties
3 required
McpToolDef
object
One tool discovered on a remote MCP server.
4 properties
2 required
ResponsesResponse
object
Non-streaming response returned when stream is false
9 properties
6 required
ImageSearchQueriesEvent
object
5 properties
4 required
ToolCallDetails
object
Details about a tool call invocation
3 properties
SearchRecencyFilter
string
Time-based recency filter for search results
ImageSearchResultsOutputItem
object
4 properties
2 required
ImageSearchResultsEvent
object
6 properties
4 required
The full machine-readable OpenAPI contract behind this narrative.
Other APIs Perplexity publishes across the network.
