Memory
Stop explaining your project every session.
Memory learns facts from your sessions over time and gives every agent the current value, on any machine.
"Our production deploy target is Hetzner Cloud, Falkenstein region."
| Fact | Value | Valid from | Until |
|---|---|---|---|
production deployed_to | Hetzner Cloud, Falkenstein | 21:14:38 | now |
One fact, one row.
"We moved production: our deploy target is now Render, Oregon region."
| Fact | Value | Valid from | Until |
|---|---|---|---|
production deployed_to | Hetzner Cloud, Falkenstein | 21:14:38 | 21:15:09 |
production deployed_to | Render, Oregon region | 21:15:09 | now |
The old value is closed, not deleted.
"Is our production deploy target still Hetzner?" No. The production deploy target is now Render, in the Oregon region. It was Hetzner Cloud in Falkenstein before.
Any agent, any machine. From the live eval, 2026-09-29.
When a fact changes, the old value closes.
- 01You state a fact
“Our deploy target is Hetzner.”
- 02The fact changes
“We moved to Render.” The old row closes, kept as history.
- 03Any session asks
Every agent on your account gets “Render”.
Tested on 24 facts that changed 38 times.
Memory blocks holding an old value
none stale
“What was true then?” answered
all 38
Store eval, 2026-09-30, against append-only and overwriting stores. A test of the mechanism, not of recall in real use. Method and sources →
Built to be trusted.
Said by you, returned by a tool, or guessed by a model, and marked so.
Your facts go everywhere. A project's facts stay in it.
Every agent and machine on your account reads it.
Your agents learn as you work.
npx @context-mode/cli connects your agents. 1,000 requests free, no card.