Research
Long-running research with persistent memory
Send documents to Pi over email or chat and build up a durable, searchable body of context. Vector memory means the agent remembers across sessions, not just within one.
Who it's for: Researchers and analysts who need continuity across many documents and days.
The problem
- Chat assistants forget everything the moment the window closes.
- Uploading sensitive documents to a hosted assistant is often a non-starter.
- Re-establishing context every session wastes tokens and time.
With openclawOS
- openclawOS stores session memory in SQLite with sqlite-vec for semantic recall, so context persists across reboots.
- Forward a paper by email or drop a link in chat; Pi reads it with its tools and returns a structured summary.
- Everything is on your hardware and your LLM account, so sensitive material stays put.
How to set it up
- 1
Install openclawOS
Start the gateway where your documents and tools live.
- 2
Pair email or chat
Pair the email channel or a messenger to feed documents to Pi.
- 3
Send material
Forward papers or links; Pi ingests them and remembers the context.
- 4
Ask across sessions
Come back days later and query the accumulated memory.
Common questions
openclawOS keeps session state in a SQLite database with a sqlite-vec vector index, so memory survives restarts and supports semantic search.
Related use cases
- A personal life-OS in your messenger Let Pi triage your health stats, calendar and notes into a single morning brief on iMessage — one agent following you across channels.
- PR review and code triage from a chat app Have Pi review pull requests, summarise diffs and surface regressions, then deliver the result to Telegram or Slack — no laptop required.
Ready to build this?
Install openclawOS, pair a channel, and put Pi to work. The whole loop is under ten minutes.