>Overview

ParalleliX CODE is a terminal REPL that behaves like a coding agent: you describe a task in plain language, and it reads your files, writes changes, runs commands, and reports results. It calls the ParalleliX inference API (api.parallelix.io/v1/chat/completions, OpenAI-compatible) for every reasoning step, so the models running your code tasks execute on network GPU nodes, not on centralized cloud infrastructure.

It is not a frontier model. It runs open-source models served by the decentralized node network, with lower quality than hosted frontier services like ChatGPT or Claude. What it offers instead: verifiable execution attribution via Proof-of-Execution, $PRLX credit billing with no per-call gas, and open-source model diversity across the node catalog.

The agent has a maximum of 30 agentic loop turns per task. Inference is paid in $PRLX credits deposited once at ai.parallelix.io/developers.