MemHub for Claude Code
Auto-capture your Claude Code sessions into MemHub team memory. At the end of
each session, an agent hook reads the transcript and saves it through the
memhub-staging MCP server, which runs tool-aware (agentic) extraction of
facts, episodes, and artifacts.
What's in here
This repo is both a marketplace and a single plugin:
.claude-plugin/marketplace.json # makes the plugin installable
plugins/memhub/
├── .claude-plugin/plugin.json # plugin manifest
├── .mcp.json # the memhub-staging MCP server (per-user OAuth)
└── hooks/hooks.json # SessionEnd → agent hook → import_conversation
Install
/plugin marketplace add XTraceAI/memhub-claude-plugin
/plugin install memhub@memhub
Then authenticate the MCP server once (the hook can't run until it's connected):
/mcp
Select memhub-staging, choose Authenticate, and approve in the browser.
How it works
- SessionEnd fires when a Claude Code session ends.
- The agent hook (a subagent with MCP access) reads the session transcript
.jsonl and calls import_conversation, passing the raw transcript records
and the session_id as the conversation_id.
- The MCP server auto-detects the Claude Code shape and routes to the
agentic ingestion path: tool-bearing events are extracted with the
agentic prompt variant (the agent is treated as a valid belief source;
facts/episodes/artifacts land in your personal team-LTM).
- Re-running the same session dedups (the
conversation_id keys a deterministic
re-import), so nothing is double-saved.
Incremental flush on commit / PR (v0.2)
Besides the SessionEnd backstop, a PostToolUse hook watches for git commit,
gh pr create, and gh pr merge and flushes the transcript-so-far in the
background (async — never blocks your session). Commits are semantic work
boundaries: flushing there makes memory available mid-session (parallel
sessions see fresh decisions minutes after each commit), shapes episodes into
work-unit narratives, and survives sessions that never end cleanly. All
triggers share one conversation_id (= session_id) and one server-side
watermark, so the full transcript is re-sent but only the delta is ever
processed — total extraction cost is the same as a single end-of-session
import.
Extra prerequisite for the flush hook: it authenticates with the
memhub CLI token (the async hook can't reuse the session's MCP OAuth), so
run once:
uvx memhub login
Without it the hook degrades silently (the SessionEnd agent hook still
captures everything at close).
Notes & trade-offs
- Auth is per-user. Nothing secret travels with the plugin — each person
authenticates the MCP server themselves via
/mcp. The first time the hook
tries to use the server you'll see a trust/permission prompt; that's by design.
- Cost. The agent hook reads the full transcript and passes it to the tool
at session end. For long sessions that's a non-trivial token cost once per
session. (A future refinement: incremental capture via a transcript watermark.)
- Requires the backend
import_conversation tool (MemHub-Backend branch
fm-feat/mcp-agentic-session-ingest) to be deployed to staging.
Configuration
To point at a different MemHub instance, edit plugins/memhub/.mcp.json
(url and oauth.clientId).
