session-sync

session-sync

A Claude Code plugin that syncs your Claude Code conversations to Effective sessions.

What it does

session-sync watches your Claude Code transcript files and uploads new conversation events (prompts, responses, code, commands, and tool output) to the Effective backend so your sessions show up in Effective.

Architecture: decoupled background daemon

The plugin never uploads on the conversation critical path. A single machine-level background daemon tails the Claude Code transcript .jsonl files and uploads deltas out of band. The only thing that runs during your session is a SessionStart hook performing a sub-50ms "ensure the daemon is running" healthcheck — it spawns the daemon detached if needed and returns immediately. The daemon polls the transcript directory every ~3 seconds, reads only the bytes appended since its last cursor, redacts secrets, and PUTs the session then POSTs the new events. Nothing in this path can slow down or block your conversation.

Install

Add this repo as a marketplace and install the plugin:

/plugin marketplace add effectiveailabs/effective-claude-plugin
/plugin install session-sync@effective

Then enable syncing with your Effective API key:

effective enable --token sk-eai-...

enable prints a consent disclosure describing exactly what is uploaded, writes config, and starts the background daemon.

You can also use the slash commands: /effective-enable, /effective-status, /effective-pause, /effective-disable.

Privacy and redaction

Transcripts include your prompts, Claude's responses, code, file contents, shell commands, and tool output. Before upload, each line is scanned and secrets matching common patterns are replaced with [REDACTED]:

• sk-... style API keys
• AWS access key ids (AKIA...)
• GitHub tokens (ghp_, gho_, ghu_, ghs_, ghr_)
• Bearer ... tokens
• PEM private key blocks
• password=, api_key=, secret=, token= style assignments

This is best-effort and not a guarantee that all sensitive data is removed. You can add your own regex patterns via the redact array in config.json.

Pause / disable

effective pause     # keep the daemon running but stop uploading
effective resume    # resume uploading
effective disable   # stop uploading and stop the daemon entirely
effective status    # show config, daemon state, and per-file sync progress

Where config and logs live

By default, under $HOME/.config/effective-sync/ (override with the EFFECTIVE_SYNC_HOME environment variable):

• config.json — apiBase, token, enabled/paused flags, extra redact patterns
• state.json — per-file sync offsets, sequence counts, cursors, last error
• daemon.pid — single-instance lock
• daemon.log — daemon activity and errors

Transcripts are read from ${CLAUDE_CONFIG_DIR:-$HOME/.claude}/projects.

Sources

The daemon watches multiple agent transcript sources and tags each uploaded batch with its source:

• Claude Code — ~/.claude/projects/**/*.jsonl (override the root with CLAUDE_CONFIG_DIR)
• OpenAI Codex — ~/.codex/sessions/**/rollout-*.jsonl (override the root with CODEX_HOME)

License

MIT © Effective AI Labs