agent-memory

name	agent-memory
description	Store, search, and manage shared knowledge through the memory service. Use for persisting information across agent sessions, semantic search over stored knowledge, and inter-agent knowledge sharing.

Agent Memory

Skill for interacting with the agentd memory service — a vector-backed knowledge store for agents.

Storing Memories

# Store a simple fact
agent memory remember "The deployment key is stored in 1Password vault 'Infrastructure'"

# With metadata
agent memory remember "OAuth tokens expire after 24 hours" \
  --created-by worker-agent \
  --type information \
  --tags auth,tokens,expiry

# Store a question for later resolution
agent memory remember "Should we migrate to async Redis client?" \
  --type question \
  --tags redis,architecture

# Store a request
agent memory remember "Need someone to review the auth middleware rewrite" \
  --type request \
  --tags review,auth

# Control visibility
agent memory remember "API key rotation schedule: first Monday of each month" \
  --visibility shared \
  --share-with ops-agent,deploy-agent

# Private memory (only visible to creator)
agent memory remember "My working hypothesis: the leak is in the connection pool" \
  --visibility private \
  --created-by debug-agent

# Reference other memories
agent memory remember "Follow-up: the Redis migration decision was made" \
  --references <memory-id-1>,<memory-id-2>

Memory Types

• information — facts, decisions, observations (default)
• question — open questions needing resolution
• request — action items or asks

Visibility Levels

• public — visible to all agents (default)
• shared — visible to creator and specified actors
• private — visible only to creator

Searching Memories

Semantic similarity search over all stored memories:

# Basic search
agent memory search "how do we handle token expiry"

# Search as a specific actor (respects visibility)
agent memory search "deployment procedures" --as-actor deploy-agent

# Filter by type and tags
agent memory search "auth" --type information --tags auth

# Limit results
agent memory search "Redis" --limit 5

# Time-bounded search
agent memory search "incidents" --since 2026-01-01T00:00:00Z --until 2026-03-01T00:00:00Z

# JSON output
agent memory search "auth" --json

Retrieving and Listing

# Get a specific memory by ID
agent memory recall <memory-id>

# List memories with filters
agent memory list
agent memory list --type question
agent memory list --tag auth
agent memory list --created-by worker-agent
agent memory list --visibility public
agent memory list --limit 20 --offset 40

Updating Visibility

# Make a memory shared with specific agents
agent memory visibility <memory-id> shared --share-with ops-agent,deploy-agent

# Make public
agent memory visibility <memory-id> public

# Make private
agent memory visibility <memory-id> private

Deleting Memories

agent memory forget <memory-id>

Health Check

agent memory health

MCP equivalents

When running as an agent with the agentd MCP server connected, the core operations are also available as tools (no CLI needed): remember (store a memory), search_memories, list_memories, and get_memory. Prefer these during autonomous execution; use the CLI for interactive/human use, visibility changes, and deletion.

remember takes content and created_by (your agent identity) plus optional memory_type (information | question | request), tags, visibility (public | private | shared), shared_with, and references.