developer-tools ai-agents code-generation memory-management

Open-source memory for coding agents, synced over SSH

GitHub - vshulcz/deja-vu: Memory layer for coding agents: search, MCP recall, auto-context, secret redaction, stats, share and sync over the session logs Claude Code, Codex and opencode already write. One zero-dep binary.

github.com

July 15, 2026

6 min read

🔥🔥🔥🔥🔥

53/100

Summary

Deja-vu is a zero-dependency binary that creates a memory layer for coding agents, allowing users to search through extensive conversation histories generated by Claude Code, Codex, and opencode. It features a rapid search capability and a recall tool that enables agents to reference past solutions efficiently.

Key Takeaways

Deja is a zero-dependency binary that creates a memory layer for coding agents, allowing users to search through gigabytes of historical data from previous conversations and design decisions.
The tool features include agent recall, auto-recall, secret redaction, and session sharing, enabling efficient retrieval and management of coding information.
Deja supports synchronization of memory between machines through export/import commands, ensuring that records are append-only and idempotent.
Installation can be done via various package managers or directly through a shell command, making it easy to integrate with existing coding agents like Claude Code and Codex.

Read original article

View story →

Coverage & discussions

Hacker News · github.com

Open-source memory for coding agents, synced over SSH ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Positive

Positives

The local storage approach is a game changer — it gives users full control over their memory, making it easy to inspect and edit.
Deja's ability to index past agent sessions in seconds has commenters buzzing about efficiency in retrieving previously solved problems.
Using a dictionary for memory retrieval is smart; it keeps things deterministic and focused on the exact details needed, unlike the chaos of semantic search.

Concerns

Concerns are raised about blindly injecting memories across devices — that could easily introduce errors and compromise separation.
Some users worry that filtering out signal from noise in memory systems is a significant challenge that hasn't been adequately addressed.