Reusable Runtime
A repo or folder can define an agent boundary, but the real value is reusing the runtime layer for models, tools, tasks, memory, services, and operations across many products.
Features
Downcity gives builders the reusable runtime layer that agent products keep rebuilding.
Design Principle
Downcity does not optimize for black-box replacement. It embeds agents inside existing engineering context so automation remains explainable, interruptible, and recoverable.
Runtime is the reuse layer; a repo is one possible boundary
Audit logs are governance assets
Path-fit beats platform migration
A repo or folder can define an agent boundary, but the real value is reusing the runtime layer for models, tools, tasks, memory, services, and operations across many products.
Agents can run as managed daemons, keep task context, execute tools, and stay reachable through CLI, Console, browser entry, or SDK surfaces.
Give agents useful tools without losing control. Runtime boundaries, permissions, logs, and operator surfaces keep execution inspectable and recoverable.
Sessions, logs, runtime state, task artifacts, and workboard snapshots make agent behavior visible instead of hidden inside one-off chats.
Plugins, skills, City, and SDK APIs let each product add capabilities without forking the whole agent backend.
A repo or folder can define an agent boundary, but the real value is reusing the runtime layer for models, tools, tasks, memory, services, and operations across many products.
Agent reliability does not come from a hidden platform. It comes from visible, auditable, and takeover-friendly configuration living in the repo.
Behavior Layer
Define personality & rules in plain English
# Personality
You are a Market Researcher.
Traits: Insightful, data-driven.
# Workspace
This repository is your workspace.
# Capabilities
## Scripts
You can run tools in ./scripts/:
- ./scripts/fetch_news.ts
- ./scripts/analyze.py
## Coding
- Use web/data tools when needed
- Write & execute custom scripts
in the ./code/ directory
Builder Decision
Close one real agent product loop on a reusable runtime, then expand it into more workflows and product surfaces.