Memory Layer
ReferenceCLI

CLI: query and briefings

These commands are the read path agents should use before making project-specific claims. They turn curated memories, activity, commits, and source evidence into answers and briefings.

CommandWritesPurpose
memory queryAudit onlyAsk a project-specific question against curated memory.
memory resumeAudit onlyGenerate a briefing for returning to interrupted work.
memory up-to-speedAudit onlyGenerate a broader new-agent project briefing.
memory activitiesNoList persisted project activity events.
memory historyNoShow memory version history, including tombstones.
memory verify-provenanceNoCheck memory source evidence against the filesystem.

memory query

Ask concrete project questions and inspect the cited memories.

memory query --project memory --question "How does the MCP server resolve projects?"
memory query --project memory --question "What changed in the docs site?" --json

Useful options vary by version, so check memory query --help; common knobs include result limits, confidence thresholds, output modes, graph-aware retrieval, and LLM-backed answer synthesis.

Good questions name the project object, subsystem, file, workflow, or decision you care about. Avoid vague prompts like “what do you know?” when you need actionable context.

memory resume

Resume is for returning after interruptions.

memory resume --project memory
memory resume --project memory --limit 20

It combines recent activity, changed memories, warnings, next actions, and checkpoint context when available.

memory up-to-speed

Use this for a new agent or a wider handoff.

memory up-to-speed --project memory
memory up-to-speed --project memory --include-llm-summary

It is broader than resume and should be run at task start when the agent lacks project continuity.

memory activities

Activities are persisted audit events.

memory activities --project memory --limit 30
memory activities --project memory --kind checkpoint --json

Use activity output to verify whether captures, curation, proposals, embedding jobs, watcher events, or checkpoints actually happened.

memory history

History shows how canonical memories evolved.

memory history --memory-id <uuid>

Use it when a remembered fact seems stale, replaced, or tombstoned.

memory verify-provenance

Verify provenance before relying on a memory that cites files or directories.

memory verify-provenance --project memory
memory verify-provenance --project memory --json

This does not prove a memory is semantically correct, but it catches missing, moved, or changed source evidence.

© 2026 Olivier Van Acker (3vilM33pl3). Memory Layer is AGPL-3.0-or-later with commercial licensing available.

On this page