Remember Conversations

Import conversation history into Neotoma memory. Supports ChatGPT JSON exports, Claude history files, Slack archive exports, and shared conversation URLs.

When to use

When the user wants to persist past AI conversations, chat transcripts, or messaging archives into durable memory — preserving decisions, commitments, and context that would otherwise be lost.

Prerequisites

Run the ensure-neotoma skill first if Neotoma is not yet installed or configured in your current harness.

Supported sources

Source	Format	Method
ChatGPT	JSON export (conversations.json)	File read
Claude	Conversation history files	File read
Slack	Archive export (ZIP with JSON channels)	File read
Shared URLs	ChatGPT/Claude share links	Web scraper MCP or fetch
Meeting transcripts	VTT, SRT, TXT, MD files	File read

Workflow

Phase 0: Verify Neotoma

Confirm Neotoma MCP is connected (call get_session_identity).

Phase 1: Identify source

1. Ask the user what to import:
   • ChatGPT export file path
   • Claude conversation history directory
   • Slack export archive path
   • A shared conversation URL
   • A transcript file
2. Detect the format from the file extension or content structure.

Phase 2: Parse and preview

1. Read the file or fetch the URL.
2. Parse conversations into a structured list: title, date, message count, participants.
3. Present a preview: total conversations, date range, highlights.
4. Ask the user to confirm which conversations to import (all, or selected by title/date).

Phase 3: Extract entities

For each conversation:

1. Store the conversation as a conversation entity with title and date.
2. Store each substantive message as a conversation_message entity linked via PART_OF.
3. Extract embedded entities:
   • Decisions: conclusions, choices, architectural decisions
   • Tasks: commitments, action items, follow-ups
   • Contacts: people mentioned or participating
   • Events: scheduled meetings, deadlines
4. Set source_file to the original filename for file-based imports.
5. Use the combined store path (entities + file) so the raw export is preserved as a source.

Phase 4: Reconstruct timeline

After importing, reconstruct a timeline of key events from the conversations:

• Decisions made (with date and conversation source)
• Tasks committed to (with assignee and deadline if available)
• People involved (with context of the relationship)

Present the timeline to the user with provenance: each event traced to a specific conversation and message.

Phase 5: Report results

Summarize:

• Conversations imported
• Entities extracted by type
• Timeline events reconstructed
• Offer follow-up queries ("What decisions did I make about X?")

Do not

• Import without user confirmation.
• Store raw conversation content in chat replies beyond what answering requires.
• Create duplicate conversation entities — check by title and date before storing.