Capped at 1,000 characters. Preview — does not persist yet.
Phase 1 preview: the toggles render but don't persist yet. Writes
land with Phase 2 via the config-api Worker. See /roadmap.
Custom instructions
How you want to be talked to.
Add a few lines that any agent will append to its default
prompt when it talks to you — preferred reply length, tone,
format, anything you'd otherwise re-explain every conversation.
Phase 2 wires this to config-api. Until then, the field renders
but doesn't persist. Capped at 1,000 characters.
Memory
See what's remembered. Delete what isn't yours to keep.
What Basal remembers about you
These are the durable facts the agent holds across conversations.
You can clear individual entries or wipe all of them.
Memory on
2026-05-08 Prefers concise responses without bullet lists unless asked.
2026-05-06 Working on the AI platform website expansion.
2026-05-02 Uses pnpm, not npm. Repo is sjolundjohn/luna-intranet.
Your activity
Last 30 days, your invocations only.
Every row is one agent invocation by you. Metadata only — never
prompt or response content. This view is private to you and AI
admins; nobody else (including managers) can see it.
When
Agent
Actor
Model
Tokens
Cost
2026-05-09 14:22
Basal
you
claude-opus-4-7
4,128
$0.082
2026-05-09 11:08
Basal
you
claude-opus-4-7
1,840
$0.037
2026-05-09 09:14
Basal
via Morning briefing
you
claude-opus-4-7
920
$0.018
2026-05-08 16:45
Basal
you
claude-opus-4-7
6,210
$0.124
2026-05-08 09:14
Basal
via Morning briefing
you
claude-opus-4-7
880
$0.018
2026-05-07 13:30
Basal
you
claude-opus-4-7
2,240
$0.045
Your spend · Basal
$3.24 / $20.00
16% used · this month
Your spend · all agents
$3.24 / $50.00
6% used · this month
Your data
Take it. Delete it. Or both.
You can export everything Luna agents have stored about you —
memory entries, audit log rows, chat-thread metadata — and you
can wipe an agent's memory of you to start over. The full export
endpoint lands with Phase 2; today the request goes through an
AI admin and turnaround is within 24 hours.