Summary: How Clawless remembers (and forgets)

The short version

Clawless remembers things about you so you do not re-explain yourself every conversation. There are two layers worth keeping separate: conversation history (the literal back-and-forth of one chat, bulky, stays inside that chat) and memory (the distilled facts pulled out of those chats over time, compact, travels with you across every conversation and agent). Memory sorts into four tiers: Pinned (you said to always remember; size-capped because it goes into every prompt), Insights (engine-inferred and surfaced routinely), General (everything else, included when relevant), and Decayed (faded; still exists, but only wakes up if current context makes it look relevant again). Memories live on your computer in a local file, and they travel to the AI provider with each message they are relevant to, which is the practical privacy rule: never save anything as memory that you would not want a provider to see. All agents share the same memory pool by default.

Core ideas

Conversation history is the literal back-and-forth of a single chat. Bulky. Lives inside one conversation. Does not travel.
Memory is the distilled facts pulled out of conversations over time. Compact. Travels across every conversation and every agent.
A quiet classifier runs in the background as you talk, watching for things worth keeping (names, preferences, decisions, ongoing projects, important dates) and writing them out as short standalone memories sorted into tiers by importance.
Tier 1, Pinned. Facts you explicitly told Clawless to always remember. Follow you across every conversation. The most load-bearing tier. Hard size cap (around 2,200 characters total) because Pinned content goes into every prompt and would otherwise silently slow and cost everything down. If you hit the cap, Clawless tells you and asks you to consolidate.
Tier 2, Insights. Classifier inferences confident enough to surface routinely. You did not type them; they appeared because the engine read your conversations. Review periodically in the Memory panel; promote the important ones to Pinned with a click.
Tier 3, General. Everything else the classifier noticed but is not sure is important. Included in conversations only when they look relevant. Do not weigh down every message.
Tier 4, Decayed. The path to forgetting without deleting. If a memory has not been touched in a long time, it ages here automatically. Still exists, still visible in the panel, but Clawless does not surface it in new conversations unless something current makes it look very relevant again. If you start working on an old project after a year, the relevant Decayed memories wake back up.
This is the difference between human forgetting and computer deleting. Humans do not delete memories; we stop accessing them. Decayed mimics that.
Three pathways for memory to get in. Classifier picks it up automatically (default; most Insights and General arrive this way). You ask explicitly (“remember that I prefer…”), confirm the candidate. You add manually in the Memory panel (Add Memory, type, save).
Memory panel. Reachable from the navigation rail on the far left. Grouped by tier with Pinned at the top. Per-memory: Pin or unpin, Edit, Delete, see source (small badge that jumps you to the source conversation), check the last-seen date.
Search. Search box at the top filters across all tiers by keyword. Type a project name; see what the engine has remembered about that project.
Last-seen date. Old dates are candidates for Decayed; new dates confirm an Insight is actually being relied on.
Sharing across agents. All agents on a single Clawless install share one memory pool by default. Tell the Assistant your name, the Writer knows it too. Right default because per-agent re-explaining would be almost as exhausting as per-conversation re-explaining. Per-agent isolation is on the roadmap.
Where memories live. On your computer, in a local Clawless data file. Not on any Clawless server. If Clawless went out of business tomorrow, your memories would still be on your computer, readable by you.
Where memories travel. With every message they are relevant to, to whichever AI provider is handling that message. That is how the agent knows your name on a fresh conversation: it comes along with the prompt.
The practical privacy rule. If there is a fact you never want leaving your computer, do not save it as a memory. Anything you want kept strictly local belongs in your own notes, not in Clawless memory.
Settings worth knowing about. Auto-extraction on or off (the classifier scanning new conversations). Re-injection interval (how often Insights are re-fed into a long conversation to keep the agent grounded; default every ten messages). Capacity-pruning notification (visibility into when General memories are pruned). Decay rules (how long an unused memory takes to age down). Defaults work for almost everyone.
Editing a memory does not change yesterday’s replies. Yesterday’s reply was generated with yesterday’s memory state. Today’s edit shapes today’s reply going forward, not the history that is already written.

What changes for you

Before this lesson, “AI remembers me” was probably a vague capability with no levers attached. Now you have the levers directly: a panel that shows every fact the engine has gathered, four tiers that explain why each fact is treated differently, three pathways for adding facts, and one clear privacy rule for deciding what should and should not go in. The most useful new habit is opening the Memory panel once in the first week and reviewing the Insights tier, both to catch any inferences that landed slightly off and to pin the one or two facts you really do not want to retype. After that, memory mostly runs itself, and the time you used to spend re-explaining yourself goes back into the work.