The agent that
learns from its work
Thoth creates skills from experience, sharpens them as it works, and remembers across sessions — backed by a cognitive memory substrate, not a stateless chat loop.
curl -fsSL https://raw.githubusercontent.com/519lab/thoth-agent/main/scripts/install.sh | bash
Skills from experience
It writes and refines its own reusable skills as it solves real tasks — capability compounds instead of resetting every session.
Cognitive memory substrate
A persistent, queryable memory layer (PostgreSQL + pgvector) keeps context, decisions, and learnings alive across runs.
Runs where you work
One agent, many surfaces: a gateway for messaging platforms, ACP for editors like Zed, a terminal TUI, and an OpenAI-compatible API.
Your models, your keys
Bring any provider — Anthropic, OpenAI, Nous, local — with credential pooling, fallback chains, and per-task model tiering.
An agent with a memory of its own
Most agents forget everything the moment a conversation ends. Thoth keeps a substrate — it observes its own runs, distills what worked into skills, and carries that forward. The longer you use it, the more it knows how you work.