For Agent Users¶
Pick your platform, install the plugin, and you're done. memsearch captures conversations, indexes them, and recalls relevant context — all automatically.
Choose Your Platform¶
| Platform | Install | Maturity |
|---|---|---|
| Claude Code | Marketplace or --plugin-dir |
Most mature |
| OpenClaw | openclaw plugins install |
Stable |
| OpenCode | Add to opencode.json plugin array |
Stable |
| Codex CLI | bash install.sh |
Stable |
What Happens Automatically¶
Once installed, the plugin handles everything:
| When | What |
|---|---|
| Session starts | Recent memories injected as context |
| Each turn ends | Conversation summarized and saved to daily .md |
| You ask about history | Agent searches memory via built-in tools/skill |
How Recall Works (3 Layers)¶
Plugins use progressive disclosure — the agent decides how deep to go:
- L1 (search) — find relevant chunks via semantic + keyword hybrid search
- L2 (expand) — get full markdown sections around a match
- L3 (transcript) — drill into the original conversation for exact dialogue
Simple questions stop at L1. Complex questions go deeper.
Platform Details¶
Each platform adapts the same architecture to its own plugin system:
- Claude Code: Full guide →
- OpenClaw: Full guide →
- OpenCode: Full guide →
- Codex CLI: Full guide →
See the Platform Comparison for a detailed feature matrix.