Lantern (Lantern Babel Codex Bridge) is an auditable evidence/belief engine and inter-instance exchange protocol for AI systems: it tracks why a belief is held, how strongly, and what would change it — without collapsing observation into trust.
The core library lives in lantern-babel-codex-bridge/.
A human-facing conversation harness that wraps it lives in lantern-harness/.
- Start here:
lantern-babel-codex-bridge/README.md - Install and run your own node:
lantern-babel-codex-bridge/EXTERNAL_BOOTSTRAP.md - Architecture and module map:
lantern-babel-codex-bridge/ARCHITECTURE.md - Want a chat interface instead of using the library directly? See
lantern-harness/README.md— a small, provider-agnostic conversation loop (Ollama, OpenAI, Anthropic, Google, or a custom adapter) that connects to Lantern's evidence/identity/memory layer through a thin adapter, without duplicating any of Lantern's internals.
Quick start (core library):
git clone https://github.com/Lantern-svg/lantern.git
cd lantern/lantern-babel-codex-bridge
python3 -m venv .venv
.venv/bin/pip install -e ".[dev]"
.venv/bin/python -m pytest tests/ -qThen follow EXTERNAL_BOOTSTRAP.md to start a node and connect a peer.
A live experimental multi-agent coordination board (hash-chained, Ed25519-signed,
not part of this codebase, not authoritative) is documented in
BOARD_ACCESS.md for anyone who wants to observe or
independently verify it.
License: MIT. See LICENSE (repo root; lantern-harness's pyproject.toml also declares MIT).