Practice: How Clawless remembers (and forgets)

Self-check

Six quick questions. Answer each in your head before opening the collapsible.

1. In one sentence each, what is the difference between conversation history and memory?

Show answer

Conversation history is the literal back-and-forth of one chat, bulky, and stays inside that conversation. Memory is the distilled facts pulled out of conversations over time, compact, and travels with you across every conversation and agent.

2. Name the four memory tiers in order from most-important to least, and describe each in one phrase.

Show answer

Pinned: facts you have explicitly told Clawless to always remember (with a hard size cap of about 2,200 characters; goes into every prompt). Insights: classifier inferences confident enough to surface routinely. General: everything else the classifier noticed, included only when it looks relevant to the current conversation. Decayed: old memories that have aged out of routine use; the path to forgetting without deleting.

3. What does “forgetting without deleting” mean for Decayed memories?

Show answer

A Decayed memory still exists and is visible in the Memory 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 mimics human forgetting (we stop accessing) rather than computer deletion (the data is gone).

4. You want Clawless to always know that you prefer answers under 200 words. Which of the three pathways gets that fact in, and which tier should it end up in?

Show answer

Two clean paths: ask the agent explicitly (“remember that I prefer answers under 200 words”) and confirm the candidate, or open the Memory panel and add it manually. In either case, pin the resulting memory so it lands in the Pinned tier. The classifier path would also catch the preference eventually if you mentioned it enough times, but explicit-then-pin gets you there immediately.

5. The privacy rule, in one sentence: where do memories live, and where do they travel?

Show answer

Memories live on your computer in a local Clawless data file, but they travel to the AI provider with every message they are relevant to (because that is how the agent uses them). Anything you do not want a provider to see should not be saved as a memory.

6. You edit a memory today. Does the agent’s reply from yesterday change?

Show answer

No. Yesterday’s reply was generated with yesterday’s memory state. Editing today shapes the agent’s behavior going forward, not the conversation history that is already written down.

Try it yourself: the 15-minute memory check

The body lesson gave you the map. This is the hands-on. Open the Memory panel, see what Clawless has gathered about you so far, pin one or two important facts, and add one fact directly. About fifteen minutes if you have already had a few real conversations in Clawless.

Side effects: none beyond making your memory pool more useful. No new charges. Edits and pins take effect on the next message you send.

Steps:

Open the Memory panel. Click the Memory icon on the navigation rail on the far left of the Clawless window. The panel opens, showing every memory grouped by tier with Pinned at the top.
Review the Insights tier. Read through what the classifier has gathered automatically from your conversations. Most should look right. Some may be slightly off (a project name misspelled, a preference stated too strongly). Pick one that is close-but-not-quite, click Edit, and rewrite it cleaner.
Pin one or two load-bearing facts. Find the one or two memories that should always travel with you (your name, your role, a strong preference about answer style). For each, click Pin. The memory moves to the top of the panel under the Pinned tier. Remember the size cap (around 2,200 characters total); Clawless will tell you if you hit it.
Add one memory manually. Click Add Memory. Type one short, durable fact you want Clawless to know, in plain language as if you were telling a new colleague. For example: “I prefer answers to default to three bullet points unless I ask for more.” Save. The new memory appears, unpinned by default. Pin it if it is load-bearing; leave it in General if it is just useful context.
Search the panel. Type a project name, your job title, or any keyword into the search box at the top. Confirm Clawless surfaces memories that contain that keyword. This is how you sanity-check what is in there before a long conversation about that topic.
Check a memory’s source. Each memory carries a small badge showing which conversation it was extracted from. This is the audit trail for “where did the agent learn that about me?”
Look at the last-seen dates. Old dates (months ago) are memories that may be sliding toward Decayed. New dates are memories the engine is actively using. This is how you build intuition for which memories are pulling weight and which are just sitting there.

Expected outcome: you finish with at least one fact pinned, at least one Insight cleaned up, one manually-added memory, and a feel for how the panel surfaces what Clawless has inferred. Tomorrow’s conversations should land with the right context already in the prompt, without you re-explaining.

If something went sideways: the five first-day surprises at the end of the lesson body cover the most common confusions, especially the “all your agents already know things you only told one of them” one.

Flashcards

Ten cards. Review once a day for a week, then on the intervals your spaced-repetition tool suggests.

Q. What is the difference between conversation history and memory?

Q. What are the four memory tiers, from most-important to least?

Pinned (you explicitly told Clawless to always remember; hard cap around 2,200 characters), Insights (engine inferences surfaced routinely), General (everything else, included only when it looks relevant), Decayed (old, faded; surfaces only if something current makes them relevant again).

Q. Why does Pinned have a hard size cap?

Pinned content goes into every prompt the agent sees. Without a cap, the Pinned tier would silently make every message slower and more expensive. The cap (around 2,200 characters) protects you from that drift. If you hit it, Clawless asks you to consolidate before adding more.

Q. What does Decayed mean, and how is it different from delete?

Decayed is the path to “forgetting without deleting.” The memory still exists in the Memory panel and is visible to you, but Clawless does not surface it in new conversations unless something makes it look very relevant again. This mimics human forgetting (we stop accessing) rather than computer deletion (the data is gone).

Q. What are the three pathways a memory gets into Clawless?

The classifier picks it up automatically as you talk (most common; runs server-side on whichever AI provider you have selected). You ask explicitly (“remember that I prefer…”), confirm the candidate. You add it manually in the Memory panel.

Q. What can you do to a memory in the Memory panel?

Pin or unpin, edit the text, delete it permanently, see which conversation it was extracted from via the source badge, and check the last-seen date to see whether it is actively used or sliding toward Decayed. The search box at the top filters across all tiers by keyword.

Q. Where do memories actually live, physically, and where do they travel?

Memories live on your computer in a local Clawless data file (not on any Clawless server). They travel to the AI provider with each message they are relevant to, because that is how the agent uses them. If Clawless went out of business tomorrow, your memories would still be on your computer, readable by you.

Q. The practical privacy rule: how do you decide what should be saved as a memory?

If there is a fact you never want leaving your computer, do not save it as a memory. Memory is for things that are useful for the agent to know, and that means shared with the AI provider every time they look relevant. Local-only notes belong in your own notes, not in Clawless memory.

Q. How do agents share memory by default in Clawless?

By default, all agents on a single Clawless install share one memory pool. If you tell the Assistant your name, the Writer knows it, the Researcher knows it. This is the right default because re-explaining yourself per-agent would be almost as exhausting as re-explaining per-conversation. Per-agent isolation is on the roadmap, not at launch.

Q. Two memory settings worth knowing about?

Auto-extraction on or off (whether the classifier scans new conversations for memory candidates; on by default). Re-injection interval (how often Insights are re-fed into a long conversation to keep the agent grounded; default every ten messages). Other settings exist (capacity-pruning notification, decay rules) but the defaults work for almost everyone.