---
type: resource
status: mixed
tags: [ai, agents, architecture]
---
# 🤖 Agent Architectures
> Patterns for building autonomous AI agents. Grounded in the [[Clawd Ecosystem]] work.

## Core components (from Clawd)
- **Core memory** — persistent `MEMORY.md` the agent reads/edits (this vault's [[Decision Log]] mirrors the idea)
- **Skills / capability registry** — browsable, editable set of agent abilities
- **MCP** servers — tool integration layer → [[MCP]]
- **Activity feed / observability** — see what the agent did ([[tiger-build]])
- **Agent control plane** — dashboard to manage the above ([[OpenClaw Dashboard]])

## Patterns to write up → [[60 - Knowledge]]
- [ ] Memory: short-term context vs. durable core memory vs. retrieval
- [ ] Skill/tool selection & routing
- [ ] Single-agent vs. multi-agent orchestration
- [ ] Human-in-the-loop checkpoints

## Related
[[Clawd Ecosystem]] · [[Agent Frameworks]] · [[MCP]] · [[LLMs]] · [[MOC - AI]]