The environment for operators running agent fleets.
Parallel agents are here. Planning, memory, and governance for them are not.
ODE is the structured layer that brings order to the growing chaos of multi-agent workflows — so you steer your agent fleets, not drift with them.
An IDE gives you tools to write and navigate code.
An ADE gives you tools to run and orchestrate agents.
An ODE gives you the environment to plan, remember, govern, and improve — across all your projects and agents, over time.
- Planning Engine — from long-term goals to today's first task. Five planning horizons, execution lanes that route work to the right executor (LLM, human, or both), tracked through completion in PostgreSQL.
- Institutional Memory — decisions, sources, and context persist across sessions. Structured retrieval across knowledge graph, episodic memory, and indexed sources — unified in a single recall layer.
- Observability — a pipeline that watches what your agents actually did. Event streams, alert rules, daily digests. You know what happened before you open the IDE.
- Governance Framework — maintain oversight without direct control. Documents stay current, context stays structured, the operator stays briefed.
- Unified Input — one surface where intent gets routed in the right direction. Think at the level of project goals, not agentic sessions.
Working subsystems built over months of daily use across dozens of parallel projects. Components releasing as open source.
Article series in progress at lastkey.agency.
This repo will grow into the full ODE platform as components are released.
⭐ Star this repo · Follow at lastkey.agency
Apache 2.0