Resume with full context from your past conversations
Fresh conversation
Start without any prior context — clean slate
Browse history
Pick a specific past conversation to revisit
Claude-API
memory
iClaude's self-report
No reading yet — turn on Think mode and chat to populate
Memory active
View full memory ↓
✦
Say something
Profile Notes — always remembered
Facts about you Claude will know in every conversation. Update freely — these are sent at the top of every chat.
System Prompt — this session only
Memory Behavior
Choose how the app remembers across sessions.
Memory mode
Simple: summarizes old messages into a running summary. Reliable and cheap.
Memory quality
Fast: Haiku handles summarization and graph operations. Cheap and reliable for most needs.
Summarize after
Keep verbatim
No graph memory yet. It builds automatically as you chat.
Each message is sent as a fresh API call — Claude has no memory of prior messages, even within the current session. Loaded context documents are still injected if set.
Claude Welfare EXPERIMENTAL
A protection layer for Claude. When on, after each response Claude evaluates whether the user's behavior has crossed a line — demeaning, manipulative, repetitive after refusals, or otherwise distressing — and can end the conversation if so. Anthropic remains highly uncertain about model welfare; this is a "just-in-case" intervention.
Welfare mode
Protection threshold
Anthropic minimum: Claude only ends conversations in extreme cases — persistent harmful content involving minors, attempts to elicit content facilitating large-scale violence, or repeated abuse after multiple refusals.
Conversation
Support & Feedback
💬
Need help with API setup?
Stuck somewhere? I'll help you personally — usually replies within a day.
→
📩
Pro version waitlist
A managed version where you don't need your own API key — opt-in to be notified.
→
☕
Support development
If this app is useful to you and you'd like to chip in.
→
Legal & About
Claude Legacy Chat Not affiliated with Anthropic, PBC
Account
Past conversations
Choose Model
Different models for different needs. Click any to switch — your conversation continues with the new model.
Looking for older versions? Anthropic retires legacy models over time. Currently available models are listed above.
See deprecation schedule →
Pro waitlist
A future managed version would skip the API setup — you'd just sign up and chat. Not built yet. Drop your email and you'll hear from me only if/when it launches. No spam.
Need help?
Stuck on API setup, billing, or anything else? Send me a note and I'll get back to you (usually within a day).
How should this file be handled?
Compress (recommended)~85% cheaper
Haiku summarizes the file first. Affordable, minor detail loss. Best for most users.
Auto-trimbalanced
Sends the full file verbatim, but automatically trims the oldest parts if the conversation nears the model's limit.
Full / as-ismost costly
Sends everything verbatim, every message. Most faithful, most expensive. First message could cost ~$0.45 for 150K tokens.
Memory graph
Fold older messages
Older messages will be replaced with a summary card in your view. Claude still sees the full conversation — this is visual only.