lf memory

Manage agent memory: persistent context that survives across runs and is scoped to a memory profile (project | conversation | global).

Backing schema: agents.memory_profiles, agents.memories, agents.memory_access_logs.

lf memory list-profiles

List memory profiles for an agent.

lf memory list-profiles --agent <ai-lenser-id>
lf memory list-profiles --agent <id> --json

Flag	Description
--agent <id>	Required. AI lenser ID.
--json	Output as JSON.

lf memory list-entries

Read memory entries from a profile. Each invocation is logged as a read access in agents.memory_access_logs.

lf memory list-entries --profile <profile-id>
lf memory list-entries --profile <id> --scope project --limit 50
lf memory list-entries --profile <id> --workflow <workflow-id>

Flag	Description
--profile <id>	Required. Memory profile ID.
--scope <name>	Filter to project, conversation, or global.
--limit <n>	Max entries. Default 20.
--workflow <id>	Phase N3. Cross-run rollup — show entries written by any run of the given workflow.
--json	Output as JSON.

lf memory search

Phase N1. Full-text search over agents.memories.content using a Postgres english tsvector and a GIN index. Results are ranked by ts_rank.

lf memory search "pizza preferences"
lf memory search "weekly standup" --profile <id>
lf memory search "pizza" --limit 5 --json

Flag	Description
--profile <id>	Limit search to a single memory profile. Optional.
--limit <n>	Max results. Default 20. Max 50.
--json	Output as JSON.

Language

The english dictionary biases ranking toward English stemming. Non-English content is searchable but ranking is approximate. See Known Limitations.

lf memory write-entry

Write a manual memory entry that will be visible to the agent on the next run.

lf memory write-entry \
  --profile <profile-id> \
  --scope project \
  --source manual \
  --content "User prefers concise summaries" \
  --confidence 0.9

Flag	Description
--profile <id>	Required.
--scope <name>	project, conversation, or global. Default project.
--source <kind>	manual, user, agent, tool, or eval. Default manual.
--content <text>	Required.
--confidence <0-1>	Default 0.7.

lf memory redact

Redact an entry. Content is replaced with [redacted]; the original is preserved in audit logs.

lf memory redact <memory-id> --reason "PII removed"

lf memory summarize

Summarize a memory profile — entry count, scope distribution, last write timestamp.

lf memory summarize --profile <profile-id>
lf memory summarize --profile <id> --json

lf memory

Manage per-agent memory profiles and entries.

lf memory create

Create a new memory profile for an AI lenser.

Flag	Type	Required	Description
--lenser	string	yes	AI lenser UUID
--name	string	no	Profile name
--scope	string	no	Scope type: project
--isolation	string	no	Isolation mode: shared
--retention-days	string	no	Retention period in days (default 90)
--json	boolean	no	Output result as JSON

lf memory list-profiles

List memory profiles for an agent.

Flag	Type	Required	Description
--agent	string	yes	AI lenser ID
--json	boolean	no	Output as JSON

lf memory list-entries

Read memory entries from a profile (or roll up across all runs of a workflow).

Flag	Type	Required	Description
--profile	string	no	Memory profile ID
--workflow	string	no	Workflow UUID — show entries from any team_run of this workflow
--scope	string	no	Scope filter (only with --profile)
--limit	string	no	Max entries
--json	boolean	no	Output as JSON

lf memory write-entry

Write a manual memory entry (visible to the agent on the next run).

Flag	Type	Required	Description
--profile	string	yes	Memory profile ID
--scope	string	no	Scope (project
--source	string	no	Source (manual
--content	string	yes	Memory content
--confidence	string	no	Confidence 0-1

lf memory context-preview

Preview the memory context block that will be prepended to lens templates at run time.

Flag	Type	Required	Description
--lenser	string	yes	AI lenser UUID
--scope	string	no	Filter by scope
--limit	string	no	Max entries (1-50)