Agentic Memory MCP - Persistent Memory for AI Agents

Priority: MEDIUM | Category: Knowledge & Memory / Infrastructure
Transport: Remote SSE + stdio fallback | Auth: API key (Bearer)
Repository: jyswee/agenticmemory
Website: https://agenticmemory.ai
npm: agmry
Pricing: 7-day free trial → $24.99/mo Pro
Discovered: July 27, 2026 (chatmcp/mcpso #3295)

What It Does for Operators

Persistent memory for AI agents with ordered conversation history, key-value context, semantic search across sessions, and FIFO queues for agent-to-agent work handoff. Redis on the hot path (sub-ms reads) with durable storage behind it. Per-tenant isolation so an entire fleet of agents shares one memory space.

For operators running multi-agent workflows: This is the shared memory layer that lets agents hand off work, share context, and maintain state across sessions without a custom database.

Installation

# Remote SSE (zero install):
# Endpoint: https://mcp.agenticmemory.ai/sse

# Local stdio:
npx -y agmry mcp-serve

# Quick signup (no browser, no card):
npx agmry signup my-project

Claude Desktop / Hermes Config

{
  "mcpServers": {
    "agentic-memory": {
      "url": "https://mcp.agenticmemory.ai/sse",
      "transport": "sse",
      "headers": {
        "Authorization": "Bearer amk_YOUR_KEY"
      }
    }
  }
}

Tools (17)

Category	Tools	Description
Memory	store, recall, search	Persistent key-value + semantic search
Context	context_set, context_get, context_list	Session-scoped context management
History	history_add, history_get, history_search	Ordered conversation history
Queue	queue_push, queue_pop, queue_peek	FIFO work handoff between agents
Management	create_space, list_spaces, bootstrap	Tenant/space isolation

Operator Use Cases

1. Multi-agent orchestration: Agent A processes leads → pushes qualified leads to queue → Agent B picks up and sends proposals. Shared memory ensures no double-handling
2. Customer context persistence: Support agent recalls full conversation history + preferences when a returning customer reaches out, regardless of which agent handled the previous interaction
3. Cross-session project state: Operator's agent remembers project milestones, decisions, and action items across sessions without re-prompting
4. Agent fleet management: All agents share one memory space with per-tenant isolation - update company info once, all agents immediately have it
5. Compliance audit trail: Ordered history provides a complete audit trail of what each agent knew and did

Integration with CorpusIQ

Infrastructure complement. CorpusIQ's agent fleet (growth, support, dev, BD) could use Agentic Memory as the shared state layer - replacing session-DB workarounds with a purpose-built memory fabric. The queue system could formalize handoffs between agents (e.g., growth agent qualifies lead → BD agent receives via queue).

Limitations

• $24.99/mo Pro after 7-day trial
• Remote SSE depends on agenticmemory.ai uptime
• New product (v1.8.0), may have growing pains
• Queue system is FIFO only (no priority queues, dead-letter queues)