agenity
A runtime that supervises fleets of parallel AI coding agents — and coaches them when they drift.
Overview
agenity is a daemon and dashboard for the operator of many AI agents. It watches every coding-agent session running across your repositories, scores each on goal, velocity, focus and end-state-proximity, and intervenes — with a coaching message — only when a session genuinely diverges from its work.
Inspired by the k9s experience for Kubernetes operators, agenity is the same idea for the parallel-AI-agent operator: one place to see every session at once, and an LLM judge that applies your own working rules as the rubric.
It also spawns and orchestrates teams of agents that talk to each other across hosts, so a fleet can divide work, hand off and verify without a human relaying every message.
What it does
Built to do the hard things well.
Live scoring
Goal / velocity / focus / end-state scored continuously per session.
Coaching, not noise
Most ticks are silent; it intervenes only on genuine drift.
Agent-to-agent mesh
Teams of agents that coordinate across hosts.
Your rules as rubric
Uses your own working guidelines as the judging criteria.
Who it's for
AI-native engineering teams
Run many agents in parallel without losing track of which ones are on-task.
Operators & tech leads
One dashboard for every session, with intervention only when divergence is real.
In the Manassa ecosystem
agenity is the layer that builds and operates the ecosystem itself — the AI workforce behind every other product.