Skills vs MCP vs RAG vs Memory: the four context mechanisms AI agents actually need
MaryamMiradi · x · 2026-11-19
Maryam Miradi breaks down four commonly confused agent context mechanisms using a production checkout 500-error scenario:
- Skills = Procedure: encode repeatable steps, rules, escalation judgment — "What should I do?"
- MCP = Access: connect to logs, metrics, databases, APIs — "What can I access or act on?"
- RAG = Documented knowledge: retrieve only relevant docs, runbooks, dependency maps — "What does the org already know?"
- Memory = Experience: surface lessons from past incidents (e.g. connection-pool exhaustion) — "What have I learned before?"
Combine them: Skill sets triage flow → MCP inspects live systems → RAG pulls documented knowledge → Memory reuses past lessons. Dumping everything into context doesn't make a better agent; production agents typically need all four.
More from coding & agent
- Anthropic details agent memory: 'dreaming' lets Claude organize its experience — daniel_mac8 · 2026-11-19
- The case for externalized memory as the real fix for continual learning — daniel_mac8 · 2026-11-19
- FT: US tech outages blamed on rush to fix bugs found by AI — nordicinst · 2026-11-19
- Claude Code overtakes Cursor and Copilot in job-post mentions ahead of Anthropic IPO — antgoldbloom · 2026-11-19
- Dev submits PR adding periodic checkpoints to GKE sandboxes so agents don't lose progress — yoimnotkesku · 2026-11-19
- AWS adds private CA support to Bedrock AgentCore Identity for connecting agents to self-hosted IdPs — AWS ML Blog · 2026-11-19