[{"data":1,"prerenderedAt":322},["ShallowReactive",2],{"navigation":3,"/features/memory":94,"/features/memory-surround":317},[4,22,39,68,81],{"title":5,"path":6,"stem":7,"children":8,"page":21},"Getting Started","/getting-started","1.getting-started",[9,13,17],{"title":10,"path":11,"stem":12},"Introduction","/getting-started/introduction","1.getting-started/1.introduction",{"title":14,"path":15,"stem":16},"Installation","/getting-started/installation","1.getting-started/2.installation",{"title":18,"path":19,"stem":20},"Workspace","/getting-started/workspace","1.getting-started/3.workspace",false,{"title":23,"path":24,"stem":25,"children":26,"page":21},"Configuration","/configuration","2.configuration",[27,31,35],{"title":28,"path":29,"stem":30},"Overview","/configuration/overview","2.configuration/1.overview",{"title":32,"path":33,"stem":34},"AI Providers","/configuration/ai-providers","2.configuration/2.ai-providers",{"title":36,"path":37,"stem":38},"Configuration Reference","/configuration/reference","2.configuration/3.reference",{"title":40,"path":41,"stem":42,"children":43,"page":21},"Features","/features","3.features",[44,48,52,56,60,64],{"title":45,"path":46,"stem":47},"Telegram","/features/telegram","3.features/1.telegram",{"title":49,"path":50,"stem":51},"Tools","/features/tools","3.features/2.tools",{"title":53,"path":54,"stem":55},"Scheduling","/features/scheduling","3.features/3.scheduling",{"title":57,"path":58,"stem":59},"Heartbeat","/features/heartbeat","3.features/4.heartbeat",{"title":61,"path":62,"stem":63},"Memory","/features/memory","3.features/5.memory",{"title":65,"path":66,"stem":67},"Skills","/features/skills","3.features/6.skills",{"title":69,"path":70,"stem":71,"children":72,"page":21},"Cli","/cli","4.cli",[73,77],{"title":74,"path":75,"stem":76},"CLI Commands","/cli/commands","4.cli/1.commands",{"title":78,"path":79,"stem":80},"Running as a Service","/cli/service-management","4.cli/2.service-management",{"title":82,"path":83,"stem":84,"children":85,"page":21},"Reference","/reference","5.reference",[86,90],{"title":87,"path":88,"stem":89},"Sessions","/reference/sessions","5.reference/1.sessions",{"title":91,"path":92,"stem":93},"Troubleshooting","/reference/troubleshooting","5.reference/3.troubleshooting",{"id":95,"title":61,"body":96,"description":310,"extension":311,"links":312,"meta":313,"navigation":314,"path":62,"seo":315,"stem":63,"__hash__":316},"docs/3.features/5.memory.md",{"type":97,"value":98,"toc":296},"minimark",[99,103,107,119,124,132,135,139,142,147,170,173,177,180,202,205,216,220,230,233,237,243,253,257,260,264,267,289],[100,101,61],"h1",{"id":102},"memory",[104,105,106],"p",{},"The agent starts each session fresh -- it doesn't inherently remember previous conversations. Memory in BabyClaw is file-based: the agent reads and writes files in its workspace to maintain continuity across sessions.",[104,108,109,110,114,115,118],{},"There are two kinds of memory: ",[111,112,113],"strong",{},"working memory"," (temporary, per-session) and ",[111,116,117],{},"long-term memory"," (extracted from conversations, persisted to files).",[120,121,123],"h2",{"id":122},"working-memory","Working memory",[104,125,126,127,131],{},"Working memory is a per-session scratchpad. The agent uses the ",[128,129,130],"code",{},"update_working_memory"," tool to jot down things it needs to track during a conversation -- file paths, intermediate results, IDs, or anything else that helps it stay on track.",[104,133,134],{},"Working memory is included in the agent's context on each turn within the session. It goes away when the session ends.",[120,136,138],{"id":137},"long-term-memory","Long-term memory",[104,140,141],{},"After conversations in the main session, BabyClaw automatically extracts notable information and saves it to daily memory files.",[143,144,146],"h3",{"id":145},"how-extraction-works","How extraction works",[148,149,150,154,157,160,167],"ol",{},[151,152,153],"li",{},"After the agent finishes a turn in the main session, memory extraction is queued",[151,155,156],{},"After a 5-minute debounce (to batch multiple messages), the extractor runs",[151,158,159],{},"It reviews the conversation and pulls out things worth remembering: decisions, preferences, context, lessons learned",[151,161,162,163,166],{},"It writes these to ",[128,164,165],{},"memory/YYYY-MM-DD.md"," in the workspace",[151,168,169],{},"Extraction avoids duplicating things already captured in existing memory files",[104,171,172],{},"The extraction is AI-powered -- the agent uses the same chat model to decide what's worth keeping. It skips transient chat, obvious implementation details, and anything already recorded.",[143,174,176],{"id":175},"what-gets-extracted","What gets extracted",[104,178,179],{},"The extractor focuses on durable information:",[181,182,183,186,189,192,199],"ul",{},[151,184,185],{},"Decisions you made (\"we're using PostgreSQL for this project\")",[151,187,188],{},"Preferences (\"I prefer dark mode\", \"don't use semicolons\")",[151,190,191],{},"Context about your life or work (\"starting a new job in March\")",[151,193,194,195,198],{},"Lessons learned (\"the deploy script needs ",[128,196,197],{},"--force"," on staging\")",[151,200,201],{},"Things you explicitly asked the agent to remember",[104,203,204],{},"It skips:",[181,206,207,210,213],{},[151,208,209],{},"Casual back-and-forth",[151,211,212],{},"Step-by-step implementation details",[151,214,215],{},"Things already in existing memory files",[143,217,219],{"id":218},"where-memories-live","Where memories live",[221,222,227],"pre",{"className":223,"code":225,"language":226},[224],"language-text","workspace/\n└── memory/\n    ├── 2026-02-18.md\n    ├── 2026-02-19.md\n    └── 2026-02-20.md\n","text",[128,228,225],{"__ignoreMap":229},"",[104,231,232],{},"Each daily file contains timestamped entries from that day's conversations.",[143,234,236],{"id":235},"memorymd","MEMORY.md",[104,238,239,240,242],{},"The workspace also has a top-level ",[128,241,236],{}," file for curated long-term memory. This is different from the daily files -- it's meant to be a distilled summary that the agent maintains over time, pulling the most important things from daily notes and organizing them.",[104,244,245,246,248,249,252],{},"The agent reads ",[128,247,236],{}," at the start of main sessions (it's part of the instructions in ",[128,250,251],{},"AGENTS.md","). Daily files are also read for recent context (typically today and yesterday).",[120,254,256],{"id":255},"memory-is-only-for-the-main-session","Memory is only for the main session",[104,258,259],{},"Memory extraction only runs after main session conversations. Other sessions (linked group chats, scheduled task runs) don't trigger extraction. This is a deliberate choice -- the main session is where personal context lives, and keeping it isolated prevents leaking information to other contexts.",[120,261,263],{"id":262},"session-startup","Session startup",[104,265,266],{},"At the start of each session, the agent reads:",[148,268,269,275,281,284],{},[151,270,271,274],{},[128,272,273],{},"SOUL.md"," -- who it is",[151,276,277,280],{},[128,278,279],{},"USER.md"," -- who you are",[151,282,283],{},"Recent daily memory files (today + yesterday)",[151,285,286,288],{},[128,287,236],{}," (main session only)",[104,290,291,292,295],{},"This gives it enough context to pick up where things left off, even though the actual conversation history is limited to the ",[128,293,294],{},"session.historyLimit"," most recent messages.",{"title":229,"searchDepth":297,"depth":298,"links":299},1,2,[300,301,308,309],{"id":122,"depth":298,"text":123},{"id":137,"depth":298,"text":138,"children":302},[303,305,306,307],{"id":145,"depth":304,"text":146},3,{"id":175,"depth":304,"text":176},{"id":218,"depth":304,"text":219},{"id":235,"depth":304,"text":236},{"id":255,"depth":298,"text":256},{"id":262,"depth":298,"text":263},"How BabyClaw remembers things between sessions -- working memory and long-term extraction.","md",null,{},true,{"title":61,"description":310},"rhVicAC9o1RKrkFDmMGJ9pqb1vMEzto3wN0A8f73ZaA",[318,320],{"title":57,"path":58,"stem":59,"description":319,"children":-1},"Periodic agent check-ins for proactive tasks.",{"title":65,"path":66,"stem":67,"description":321,"children":-1},"Extending the agent with skills from ClawHub or your own custom skills.",1771846894211]