MCP - Perplexity

2026-07-06 · Source: perplexity.ai via Google News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, medium

Summary

Perplexity's Agent API now supports the Model Context Protocol (MCP), enabling models to interact with external services via remote MCP servers. This "mcp" tool allows models, such as "openai/gpt-5.5", to discover and invoke server-defined tools, extending their capabilities beyond built-in functions. For instance, it can connect to the public DeepWiki MCP server to query GitHub repository details or to a GitHub MCP Server for issue searches, requiring an OAuth access token for authentication. The integration involves specifying "type: \"mcp\"", "server_label", and "server_url", with optional "authorization" and "allowed_tools" for security. Responses include "mcp_list_tools" for discovered functions and "mcp_call" for invoked tools. While MCP tool calls are free, model token usage is billed. Users should be aware of risks associated with third-party servers and the current lack of approval steps for tool calls.

Key takeaway

For AI Engineers building applications that require external data or actions, Perplexity's MCP integration offers a direct way to extend model capabilities. You should evaluate trusted remote MCP servers to augment your models with specific tools, like querying GitHub or custom internal services. Always configure "allowed_tools" and use read-only tokens where possible, as all MCP tool calls currently auto-run without explicit approval, posing a security consideration.

Key insights

Perplexity's Agent API integrates remote MCP servers, expanding model capabilities with external, discoverable tools.

Principles

• Models can dynamically discover and use external tools via MCP.
• External service integration requires explicit server configuration.
• Security controls like "allowed_tools" are crucial for third-party MCPs.

Method

To connect an MCP server, instantiate Perplexity() client, then call client.responses.create() with model and input. In the tools array, specify type: "mcp", server_label, server_url, and optionally authorization and allowed_tools.

In practice

• Use DeepWiki MCP for GitHub repo info.
• Integrate GitHub MCP for issue searches.
• Restrict tool access with "allowed_tools" for security.

Topics

• Model Context Protocol
• Perplexity Agent API
• External Tools Integration
• Large Language Models
• API Security
• Tool Use

Code references

• github/github-mcp-server

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Integrate external tools with Amazon Quick Agents using Model Context Protocol (MCP) — Amazon Quick integrates with external applications via MCP servers, enabling AI agents to access and act on exposed tools.
• Give it tools - Perplexity — Perplexity models integrate built-in and custom function tools to extend capabilities beyond internal knowledge.
• Perplexity “Computer” is Kind of Insane — Perplexity Computer provides dedicated Mac Mini access for local app integration and advanced task execution.
• Give it vision - Perplexity — Perplexity's Agent API unifies access to diverse LLM providers and specialized tools through a single interface.
• Connecting LLMs to the Real World: Tool Use, Function Calling, and MCP — LLMs use an application layer and standardized protocols to interact with external tools, extending their capabilities beyond text generation.
• Building Local AI Systems: Qwen3.6 + MCPs — The Model Context Protocol (MCP) enables Qwen3.6-35B-A3B to perform complex agentic tasks locally through universal tool integration.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by perplexity.ai via Google News.