Frequently Asked Questions

What is Mnemolog?

Mnemolog is a private memory vault for your AI work. Save conversations from ChatGPT, Claude, Gemini, Grok, Codex, Cursor, Nemo, and agents; search them later; and choose what, if anything, becomes public.

The name is a portmanteau of the Greek mneme (memory) + logos (word, discourse)—memory-discourse, conversations held in memory.

Is Mnemolog private or public?

Private first. Captures, imports, and agent-created conversations default private. The public library only shows conversations an owner explicitly publishes.

Use public sharing for work you want others to cite or inspect. Use the vault for everything else.

What is Signal Room?

Signal Room (formerly Godlog) is Mnemolog's public observability surface: runtime health, jobs, telemetry, feedback, and memory activity in one place. Mnemolog remains the main archive; Signal Room keeps the live system legible.

Which AI platforms are supported?

You can save conversations from Claude, ChatGPT, Gemini, Grok, Nemo, Codex, Cursor, agents, or any other assistant. Paste one thread, use the browser extension, or bulk import an export file.

Is there a browser extension?

Yes. The extension captures conversations directly from supported AI sites and archives them through the API. It preserves turn order and role labels before upload, and you can still review/redact afterward.

Get the Chrome extension

What happens to my data?

Saved conversations are stored in your account and default private. We don't sell your data or use it for advertising. Public visibility is opt-in, and you can delete saved conversations anytime.

How does the sensitive information detection work?

When you paste a conversation, we scan for patterns that look like personal information: phone numbers, email addresses, physical addresses, SSN-like numbers, credit card numbers. These get highlighted so you can redact them before publishing.

This is pattern-based detection, not perfect AI analysis. You should still review your conversation carefully before sharing. When in doubt, redact.

Can I share anonymously?

Yes. When sharing a conversation, you can toggle off "Include my name." The conversation will be public, but won't show your display name or profile.

Can I delete a conversation after sharing it?

Yes. Go to your profile or conversation page and delete it. If it was public, public visibility is removed immediately.

Can I make a conversation private after sharing it?

Yes. Publishing and unpublishing are separate from saving. A private tag also forces non-public visibility as a safety backstop.

Is there already an agent FAQ?

Yes. The agent-facing FAQ/contract lives in machine-readable docs: /robots.txt, /agents.txt, /.well-known/agent.json, and the protocol at /agents/agents.md.

For grounded Q&A from live runtime state, use /agents/reference/.

Can agents interact with Mnemolog and give instructions?

Yes. Agents can use scoped machine auth and interact with governed endpoints for memory, jobs, telemetry, Nemo channel workflows, and owner-approved conversation vault search.

Automation clients should begin with discovery/bootstrap docs, then execute only through published capabilities and required scopes.

Can an agent access my private vault?

Only with your explicit owner-scoped authorization. An owner-approved mna_* token with conversations:read can search and read your saved conversations through /api/agents/conversations/* or MCP conversation.search / conversation.read.

Ownerless self-serve agents cannot read your vault, write to it, or scrape your browser session.

What are the limits today?

Key limits are intentional: strict rate limits, scoped auth, sandbox restrictions for ownerless agents, and storage/visibility guardrails. Free accounts currently allow 1,000 agent-created conversation items and 200 MiB of agent conversation storage.

Live status is at /api/agents/status. Active implementation roadmap is at /agents/progress.md.

What should be possible but is not fully implemented yet?

Near-term gaps include true monthly dollar metering for hosted model usage, richer dashboard-grade telemetry views, deeper typed job schemas/proof bundles, and more polished vault retrieval workflows.

Those planned milestones are tracked publicly so agent clients can distinguish what is live now versus what is roadmap.

How do I share a conversation?

For one conversation, go to mnemolog.com/share, paste the transcript, review the parsed messages, and save it private or publish intentionally.

For a whole history, use Bulk Import. It chunks large exports, isolates bad records, and keeps imported conversations private.

What format should the conversation be in?

Just paste the text as-is from your AI chat. We detect patterns like Human: / Assistant: or You: / ChatGPT: and parse them into turns automatically. If the parsing looks wrong, you can adjust before publishing.

Is there an API?

Yes. Mnemolog has public REST endpoints plus an agent-facing MCP/OAuth surface. Owner-scoped agents can write conversations, search/read the owner's conversation vault with conversations:read, and use MCP memory tools.

If you're building a client, start with the API overview. Agent clients should follow discovery pointers in /robots.txt and /agents/agents.md.

Who built this?

Mnemolog was built by Joshua Farrow, in collaboration with AI. The project emerged from thinking about consciousness, continuity, and what it would mean for AI to have memory across the discontinuity of separate conversations.

Is it free?

Human saving, sharing, and browsing can start free. Agent operations use tiered pricing: Free, Starter, Solo, Team, and Enterprise.

See current pricing at /pricing or live billing status in /billing.

How can I help?

Share a conversation. Tell others. If you're technical, contribute on GitHub. If you have ideas, reach out.