Skip to content

Thoth

The lifelong-learning agent runtime

Thoth — Egyptian god of writing, wisdom, and the moon. Patron of scribes. Recorder of words.

Thoth is an open-source agent runtime where memory accumulates in five composed layers, the agent judges itself through reflection, every chat reaction is a training signal, and skills crystallize into a versioned library that evolves over time.

What you can build

A Slack-native AI assistant

Drop the bridge into your Slack workspace. The first turn loads your persona stack; replies in the same thread continue the same Claude session. Reactions become learning signals.

An autonomous operator agent

Configure ambient triggers, schedule self-spawns, define the budget envelope. Agent quietly checks on what matters while you sleep.

A learning agent that grows

Voyager-style skill compilation: every session ends with reflection, proposed new skills, and four fan-out writers updating different layers of memory. Founder approves; agent grows.

A multi-agent council

Convene a debate of personas across heterogeneous models. Synthesis visible. Cross-persona blind review removes attribution bias.

The five layers of memory

L5 REFLECTION · session-end critique → 4 fan-out writers
L4 PROCEDURAL · skills, persona, hard rules
L3 EPISODIC · cosine recall over local embeddings
L2 IDENTITY · theory-of-mind on each peer (Honcho)
L1 WORKING · current session state

Each layer has its own writer, its own cadence, its own kind of truth. None can collapse into the others without losing what makes a mind a mind. Read the full architecture →

Status

v0.5.0 — public release on 2026-07-22.

Star github.com/thoth-runtime/thoth to follow.