Skip to content

Back to the catalog

Pal-AI-Lab/Cortico
github.comrepo

Pal-AI-Lab/Cortico

Cortico is an open-source agent harness built around an event stream, designed for autonomous, continuously running agents that handle mixed real-time input across chat platforms, live games, and even physical environments. It provides modular LLM provider components (via an internal Responses protocol), unconstrained context/memory management, an isolated extension system called Cortico World, and a WebUI, with a companion AI-assisted extension creator (Cortina) that lets non-developers build extensions through conversation. A reference use case is cortico-world-vtuber, which drives a Live2D model via VTube Studio with streaming TTS and OBS overlay integration.

TL;DR Open-source event-driven agent harness for building persistent AI personas, VTubers, and companion bots with an extensible World/plugin system.
Open original ↗
Document

How to use it

Requires Node 22+. First launch bootstraps a default deployment named mybot (reference bot cormini, terminal chat enabled) and serves a web console at http://127.0.0.1:7788/.

corepack pnpm install
pnpm start

Each deployment is an isolated config directory under deployments/. Add or select deployments by name:

pnpm start --new
pnpm start second

The launcher scripts (pnpm start, ./start.sh, start.bat) auto-install missing dependencies, rebuild web console assets when stale, list deployments for interactive selection, and can restart processes from the console. Onboarding is three steps in the console: configure a model endpoint, inspect active Worlds, adjust the system prompt, then "Say hello" to have the bot open the conversation. Endpoints, credentials, and runtime parameters are all editable from the console; changes apply on the next request (or next start for a managed runtime).

Contributor verification, before submitting a PR:

pnpm test
pnpm run typecheck

For frontend changes, also pnpm typecheck:web and pnpm build:web. CONTRIBUTING.md covers code-separation rules and AGENTS.md the review checklist. AI-assisted contributions are accepted only if the author can explain all submitted logic.

Two things the summary understates

The VTuber reference lives outside the main tree under a different licence. cortico-world-vtuber is released under AGPL-3.0 + CLA, so it sits in a separate repository rather than the Cortico main tree. Beyond what the summary lists, it also times subtitles with a forced aligner (not just streaming TTS + OBS). If you plan to build on or redistribute the VTuber World, the copyleft + CLA terms are the thing to check first — the main Cortico repo's licence is not stated on this page.

Cortina is a separate repo with a specific workflow and a design thesis. You clone it from https://github.com/Pal-AI-Lab/Cortina, open the workspace in any coding agent, and drive extension development or adaptation through conversation; finished extensions are published to npm to share. It is built on the TINA ("There Is No App") Spec proposed by Pal AI Lab — the stated framing is "language is the code, the agent is the runtime." That is a design stance to weigh, not just a convenience tool.

Limits

This is a README. It states no benchmarks, no scale figures, no supported-model list beyond "various upstream LLM APIs and locally deployed models via the internal Responses protocol." The architecture is described as "four distinct layers" but the layers are not enumerated in the fetched text — consult the linked Architecture doc before committing. The main-repo licence is unstated here; only the VTuber World's licence (AGPL-3.0 + CLA) is given.

Organisations named here

  • Pal AI Lab

Technologies named here

  • Cortico
  • Cortico World
  • cortico-world-vtuber
  • Cortina
  • VTube Studio
  • OBS
  • WebUI
  • Node.js

Techniques named here

  • TINA Spec

Related in agents