Perplexity AI

2026Perplexity AIopen source

Numbat

v0.1.2
AI Security
Agent Security

Numbat is an open-source security framework designed to provide visibility, detection, and response capabilities for AI agents running on local endpoints. It integrates with agent harnesses to monitor activity, enforce security policies via pre-action hooks, and perform forensic reconstruction of agent sessions.

Real-time monitoring via local hooks, plugins, and OTLP/HTTP log exporters
Endpoint-local detection using Common Expression Language (CEL) and custom YAML rules
Optional pre-action blocking to prevent high-risk agent behaviors
version0.1.2
regionUnited States
2026Perplexity AIopen source

Lily

featured
Inference Engine
Apple Silicon

Lily is a high-performance, lightweight inference engine written in Rust and optimized specifically for running the Qwen3.6-35B-A3B model on Apple Silicon. It utilizes custom Metal kernels to bypass standard frameworks like PyTorch or MLX, providing a minimal OpenAI-compatible API for greedy text generation.

Custom Metal kernels for Apple Silicon optimization
Minimal OpenAI-compatible chat completions API
Fixed-block attention and GQA packing for high-speed decoding
pricingFree (Open source)
regionUnited States
2026Perplexity AIpaid

Perplexity Computer

v1
Agentic AI
AI Agents

Perplexity Computer is an AI-powered assistant designed to enhance productivity by integrating with various applications and providing advanced features. It offers seamless interactions across multiple platforms, enabling users to perform tasks efficiently.

Integration with email and calendar apps (Gmail, Google Calendar, Outlook)
Enhanced memory for personalized responses
New image generation models (Nano Banana and Seedream 4)
version1
regionUnited States
2026Perplexity AIfree

Perplexity CLI

v0.2.2
CLI
Developer Tools

Perplexity CLI is a command-line interface tool that brings Perplexity's powerful search and content extraction capabilities directly into your terminal. Designed for developers and coding agents, it provides grounded web search results and clean page content, outputting data in JSON for seamless integration with automation workflows.

Grounded web search results delivered in machine-readable JSON
Clean page content extraction with metadata and optional raw HTML
Support for advanced search filters including domains, recency, and date ranges
version0.2.2
regionUnited States