Loading…
nanoiqqi is an ultra-lightweight agent framework built around a small, readable agent loop, a smart router, and background subagents — with a built-in Brain Office / IQQI Office UI that visualizes activity in real time.
The core agent code (loop, tools, routing, subagents) is kept intentionally compact so it is easy to audit, research-friendly, and fast to start and iterate on.
nanoiqqi/agent/loop.py — input, context, LLM call, tools, subagents, output. Built-in IQQI Office UI (Brain Office) shows tools, subagents, and activity in real time. Native skills include LeadsMx for Mexico DENUE/INEGI.
Picks models by capabilities, cost, and latency. One router for main agent and subagents.
Handle long-running tasks in the background; the main agent gets a summary back.
Built for people who want a small, readable codebase — not another black box.
A few rules that guide how nanoiqqi is built.
Agent loop and tools you can read in an afternoon.
Transparent routing, tools, and skills.
OpenRouter, vLLM, Anthropic, OpenAI, and more.
Skills, MCP, and channels you can extend.
nanoiqqi is part of the IQQI ecosystem — the same team behind iqqi app. We keep the core intentionally small so you can self-host, fork, and own your automation.
Clone the repo, open an issue, or explore the codebase and make it your own. MIT License.