Cookbook

How to share memory across opencode, dashboard, and scripts

One workspace, five surfaces, zero sync code. A fact stored in opencode appears in the dashboard, the SDK, the CLI, and your embedded SaaS — immediately.

Memory is centralized, access is distributed

Memory is stored once per workspace. Every surface that uses the same API key shares the same memory pool. No syncing, no replication.

Five surfaces, one memory pool, zero sync code

A day in the life

Morning — opencode

> Our refund policy is 30 days, not 60. Update the docs.

Got it. I'll remember that our refund policy is 30 days.
[Memory: stored as knowledge atom]

Afternoon — dashboard

Visit /infra/dashboard/memory
Search: "refund policy"
Result: "Refund policy is 30 days" — stored this morning via opencode

Evening — Python script

import asyncio
from cadreen import Cadreen

async def main():
    cadreen = Cadreen()
    results = await cadreen.memory.search("refund policy")
    print(results[0].content)  # "Refund policy is 30 days"
    print(results[0].source)   # "opencode"

asyncio.run(main())

Next day — CLI

$ cadreen memory list --limit 5

Recent memories:

  Refund policy is 30 days
  Source: opencode
  Stored: 2026-06-21 09:15 UTC
  Type: knowledge_atom

Note

Memory is per-workspace, not per-surface. Source tags ("opencode", "sdk", "cli") let you filter by origin, but knowledge flows freely.

Write memory from any surface

SDK — remember a fact

await cadreen.memory.remember({
  content: "Our Stripe account ID is acct_12345",
  type: "fact",
});

CLI — add to memory

cadreen memory add "Our Stripe account ID is acct_12345"

Both of these write to the same memory pool. The fact is immediately available on all surfaces.

Note

Next: How to enable GDPR, HIPAA, and PCI-DSS compliance in one call — pre-built policy templates.