PopMCP
perplexity_cancel_Stop a batch.
Perplexity MCP server for your AI
Ask a question and get an answer with its sources, hand off a research task and pick up the report later, or pull ranked web results with the page text, all from the AI you already use.
PopMCPPerplexity
What changed in EU battery regulation this month? Cite every claim and tell me what the answer cost
Supported Perplexity tools
Every current Perplexity API is covered: the Agent API, Search, Decisions, Embeddings, Batches and Skills, plus the Enterprise usage and audit endpoints. A few tools do a whole job at once: ask one question, run long research and wait for it, or run a whole batch from a list.
38 tools in the full catalog: 29 read and 9 write. 30 load by default, and the rest are found with tool search.
perplexity_create_Start a batch of Agent API requests from an uploaded file, at batch pricing and finishing within 24 hours.
perplexity_delete_Delete a batch file for good.
perplexity_download_Read a batch's results or errors in pages of lines.
perplexity_get_Read one batch (batch_...): status, request_counts, and output_file_id and error_file_id once it is done.
perplexity_get_Read one batch file's details (file_...): name, size, purpose and when it expires.
perplexity_list_List this organization's batches, newest first, with status and request counts.
perplexity_list_List batch input, output and error files.
perplexity_run_Write a batch file, upload it and start the batch.
perplexity_upload_Upload a batch input file without starting it.
perplexity_askAsk a question and get a written answer from live web research, with citations and the cost.
perplexity_cancel_Stop a running Agent API response.
perplexity_create_Run one Agent API request with full control: effort or model, instructions, tools, skills, structured output, multi-turn chains.
perplexity_download_Read a file an Agent API run produced in its sandbox.
perplexity_get_Read one Agent API response by id (resp_...), with its answer, citations, status and cost.
perplexity_get_Pick up a research run by id, waiting if it is still going.
perplexity_list_List the files an Agent API run produced in its sandbox.
perplexity_list_List the model ids the Agent API accepts, such as openai/gpt-6-luna or anthropic/claude-sonnet-5-5.
perplexity_researchRun a long research task in the background and wait for the report.
perplexity_add_Add a new revision of a custom skill.
perplexity_create_Create a custom skill from the text of SKILL.md or a link to a zip.
perplexity_delete_Delete a custom skill for good.
perplexity_get_Read one custom skill.
perplexity_get_Get a short-lived link to download a custom skill as a zip.
perplexity_list_List the revisions of a custom skill, newest first.
perplexity_list_List the custom skills in this Perplexity Project, newest first.
perplexity_create_Embed chunks of documents so each chunk knows its document.
perplexity_create_Turn texts into embedding vectors for semantic search or clustering.
perplexity_evaluate_Ask named questions about some content and get calibrated probabilities instead of prose: yes or no (noul), one of your options (choice), or a level (score).
perplexity_search_Ranked web results with extracted page text, and no generated answer.
perplexity_orchestrator_Private preview.
perplexity_orchestrator_Private preview.
perplexity_orchestrator_Private preview.
perplexity_orchestrator_Private preview.
perplexity_orchestrator_Private preview.
perplexity_get_Enterprise only, free: organization usage of Perplexity as a time series for one dataset (credits, queries, active members, connectors and more).
perplexity_get_Enterprise only, free: daily usage per member, keyed by email, for credit_usage or query_volume.
perplexity_list_Enterprise only: the organization's audit events, newest first, at most 30 days per call.
Connect Perplexity to Claude, ChatGPT and Claude Code
Authorize once, then an endpoint your client signs into. Nothing to deploy, and nothing to install on the Perplexity side.
Open Perplexity in PopMCP, click Authorize with Perplexity, sign in, choose the API organization to bill and approve. There is no developer app and no callback URL to set up. You need to be an admin of that organization, and it needs API credits. Prefer an API key? It goes under More options.
How you connect
your-perplexity-valueAPI key (instead of signing in)
••••••••••••3f7aAnalytics API key (Enterprise)
••••••••••••3f7aAuthorize PopMCP on PerplexityAuthorize
Questions and searches change nothing in your account, so they work on a read-only connection, but they do spend your Perplexity credits. Changing skills, files or batches and using connectors wait until an owner or admin allows write actions in PopMCP, and new connections start read-only. The connection bills one API organization and stays inside Perplexity's own rate limits.
Reach Perplexity
Reach other connectors
Teammates on read-only
Drop the PopMCP URL into Claude, ChatGPT, Claude Code, Codex or any other supported client. It signs in through PopMCP with OAuth, so your Perplexity credential never leaves the server.
app.popmcp.com/mcp
Claude wants to connectApproveAsk for something harmless before the agent changes anything. Writes run as soon as it calls them, so confirm it reads the right account first.
What changed in EU battery regulation this month? Cite every claim and tell me what the answer cost
perplexity_askok
Then just ask for the work
“What changed in EU battery regulation this month? Cite every claim and tell me what the answer cost”
- ask
“Research the top ten competitors in home espresso machines and write me a report with sources”
- research
- get_research_result
“Find recent pages on nature.com about solid state batteries and give me the links with the key passage from each”
- search_web
“For each of these 200 companies, find the headquarters city and the CEO, and tell me when the batch is done”
- list_models
- run_batch
- get_batch
- download_batch_file
“Go through these support tickets and tell me how likely each one is to be a billing problem”
- evaluate_decisions
Use Perplexity from the assistant you already open
One hosted endpoint, registered once per client. The same tools and the same permissions, whichever window the request comes from.
Claude DesktopCustom connector
Add the PopMCP URL under Customize → Connectors and the tools appear in every chat.
Claude WebCustom connector
The same endpoint, registered in the browser. Nothing to install on the machine you are working from.
ChatGPTCustom connector
Register PopMCP as a custom connector, then let ChatGPT read and update the account.
CodexDesktop & CLI
Wire the endpoint into Codex or the Codex CLI when a task needs live business context, not a fixture.
OpenCodeCoding agent
Register the remote MCP server in OpenCode's config for terminal-first and scripted agent runs.
ZCodeDesktop app
Add PopMCP as an MCP server in Z.ai's desktop app and work the account from the same window.
Explore other MCP servers
Every one of them runs on the same hosted endpoint, with the same controls.
FAQ
Perplexity MCP server questions
What teams ask before putting Perplexity behind an AI client.
Do I need a Perplexity API key or developer app?
No. You open Perplexity in PopMCP, click Authorize with Perplexity, sign in and approve. Perplexity's sign-in works with no app registration, so there is nothing to create first. If you prefer not to sign in, an API key from the API portal works instead, under More options.
Does my Perplexity Pro subscription cover this?
Treat them as separate. The API bills an API organization's credits, which are bought in advance on the Billing page of the API portal. Check that page shows a balance before you start, because keys are blocked when credits run out.
What does it cost to use?
PopMCP adds nothing to Perplexity's own prices. Each call is billed to your API credits: the Search API costs $5 per 1,000 requests, and Agent API runs bill per token and per tool call. Each answer reports what it cost, so you can ask the AI to tell you.
Do questions cost money on a read-only connection?
Yes. Read-only means the AI cannot change anything in your account or act in other services through Perplexity. Asking a question or searching changes nothing there, so it is allowed, and it still spends credits. Connector and remote MCP tools are refused on a read-only connection, because they can act elsewhere.
What happens with research that takes several minutes?
It runs in the background on Perplexity's side. The AI waits for a while and, if the report is not ready, gets an id to check again later. Checking is free apart from counting against the rate limit, and it does not start or bill the task a second time.
Can it use Sonar models?
Perplexity ended support for Sonar Chat Completions on September 27, 2026 and recommends its Agent API, so that is what the main tools use. The Sonar tools are still reachable for older work, but they are not loaded by default, and Perplexity says background Sonar requests are no longer supported.
What do the Analytics and audit log keys do?
They are for Perplexity Enterprise organizations. The Analytics API key lets the AI read your organization's usage, and an audit log API key lets it read audit events. An org admin creates them in the Perplexity web app, and you add them under More options. Each key works only for its own data.
Do I have to run or host the Perplexity MCP server myself?
No. PopMCP hosts it and gives you a Streamable HTTP MCP URL. There is nothing to deploy, no callback URL to register, and no infrastructure to keep running or patch.
Which AI clients can use this MCP server?
Any client that supports a remote Streamable HTTP MCP server: Claude Desktop, Claude Web, Claude Code, ChatGPT custom connectors, Codex and the Codex CLI, OpenCode and Z.ai's ZCode, as well as editors such as Cursor, Windsurf and Zed. Your client signs in to PopMCP with OAuth; the provider credential never leaves the server.
How do I connect Perplexity to Claude Code?
Connect Perplexity in PopMCP first, then run claude mcp add --transport http popmcp https://app.popmcp.com/mcp in a terminal. Inside Claude Code, run /mcp and sign in to PopMCP in the browser window it opens, ticking Perplexity on the consent screen. The Perplexity tools are then available in that session, and no Perplexity credential is stored on your machine.
Why do only 30 of the 38 tools load by default?
Loading 38 tool definitions into every conversation burns context before the work starts. PopMCP loads a curated 30-tool set covering the everyday workflows and keeps the rest searchable, so your AI pulls a specialist tool only when a task needs it. Every tool stays searchable, and on Scale the owner can load the full catalog for a teammate's connection.
What is the difference between a tool and a trigger?
A tool is something your AI calls, and there are 38 here. A trigger is something PopMCP watches for you: 3 Perplexity events that can start work on the PopMCP side. Triggers never enter your AI client's context, and none of them ask you to paste a callback URL into Perplexity.
Connect Perplexity to your AI in minutes
Launch a hosted MCP endpoint for Perplexity, keep tool access under control, and start testing real workflow prompts right away.