Ecosystem: claude-mem + external context engines (ByteRover, etc.)

[alessandropcostabr]

Observation

claude-mem handles temporal memory (what happened in past sessions). But many users also use structural context engines that manage project-level knowledge trees (architecture docs, conventions, domain knowledge).

These are complementary:

• claude-mem: "What did I do yesterday on the auth module?"
• Context engine: "What are the auth module's design principles?"

After 23 days of running claude-mem alongside a context engine (ByteRover), we found they stack well:

Layer 1: Temporal memory (claude-mem)

• Observations from past sessions
• Session summaries
• Semantic search for relevant past work

Layer 2: Structural context (external engine)

• Project architecture docs
• Coding conventions and standards
• Domain knowledge trees

Combined workflow

SessionStart:
  1. claude-mem injects recent/relevant observations
  2. Context engine injects project knowledge tree
  3. Claude receives both temporal AND structural context

UserPromptSubmit:
  1. claude-mem: semantic search for relevant past work
  2. Context engine: relevant project docs for this query
  3. Combined context = memory + knowledge

Market landscape

Several projects address the AI agent memory space:

• Mem0, Zep, LangMem: established memory platforms
• Letta AI: stateful agents with persistent memory
• Hindsight: MCP-compatible memory server (human-like patterns)
• OpenTimelineEngine: shared Claude Code + Codex memory
• Memorix: multi-platform cross-agent memory via MCP

claude-mem's advantage is tight integration with Claude Code's hook system — it captures tool usage in real-time without requiring explicit save actions.

Questions for the community

1. Has anyone integrated claude-mem with external context tools?
2. Would a plugin/hook interface for "context providers" be useful?
3. Should claude-mem have a way to consume external context alongside its own observations?
4. What other memory/context tools are people using alongside claude-mem?