graph TD
    A[Agentic Memory] --> B[Markdown - source]
    A --> C[SQLite FTS5 - derived]
    A --> D[14-Phase Search Pipeline]
    B --> E[.md files - Git-ready]
    C --> F[Temporal Knowledge Graph]
    D --> G[CQRS + CRDT Multi-Agent Sync]
    A --> H[25 MCP tools]
    A --> I[55 cron scripts -> 1 scheduler]
    A --> J[9 hooks]
    A --> K[Python SDK + TypeScript SDK + REST API]

from agentic_memory import MemoryClient

mc = MemoryClient()
mc.save("User prefers dark mode", category="preferences")
results = mc.search("dark mode")
for r in results:
    print(f"[{r.score:.2f}] {r.content}")

from agentic_memory import AgentMemory

coder = AgentMemory(agent_id="coder")
coder.save("Frontend uses React with TypeScript")

designer = AgentMemory(agent_id="designer")
designer.save("Brand colors are #FF5733 and #33FF57")

# Add to your MCP config
{
  "agentic-memory": {
    "command": "agentic-memory-server"
  }
}

agentic-memory api --port 9878
curl http://localhost:9878/api/v1/search?q=dark+mode

retention = sigmoid(w_acc × access + w_surp × surprise + w_imp × importance + w_fit × fitness - w_rec × recency - bias)

agentic-memory/
├── agentic_memory/              # Python SDK (pip installable)
│   ├── client.py                # MemoryClient (save/search/CRUD)
│   ├── temporal.py              # TemporalKG
│   ├── kg.py                    # KnowledgeGraph
│   ├── integrations/            # LangChain + CrewAI adapters
│   └── models.py                # 8 typed dataclasses
├── search/                      # 14-phase search pipeline
│   ├── orchestrator.py          # Main pipeline (2,825 LOC)
│   ├── scoring.py               # RRF, temporal decay, KG boost
│   ├── rerankers.py             # Cross-encoder, ColBERT
│   ├── chunk_index.py           # Semantic chunking
│   └── synthesis.py             # Answer synthesis
├── save/                        # Write path
│   ├── pipeline.py              # Saga-wrapped save
│   ├── backlinks.py             # Wiki-style backlinks
│   └── post_save_hooks.py       # Post-save operations
├── infra/                       # Infrastructure
│   ├── db.py                    # Connection pool + WAL
│   ├── write_journal.py         # CQRS write journal
│   ├── embedding_search.py      # Semantic embeddings
│   ├── reranker.py              # Neural reranker
│   ├── vector_store.py          # ANN index abstraction
│   ├── api_server.py            # REST + WebSocket
│   └── cache.py                 # Multi-level caching
├── knowledge_graph/             # KG extraction + search
├── kg/                          # Temporal KG + analytics
├── crdt/                        # Field-level CRDT merge
├── fact/                        # Fact extraction + temporal
├── background/                  # Daemon + worker + circuit breaker
├── cron/                        # 47+ cron jobs + consolidated scheduler
├── hooks/                       # 6 lifecycle hooks
├── migrations/                  # 57 reversible migrations
├── eval/                        # 363 test files, 5,703+ test functions
├── ts-sdk/                      # TypeScript SDK
├── mcp_*.py                     # 31 MCP modules
├── mcp_health.py                # System health MCP tool
└── dashboard.py                 # Streamlit observability

pip install agentic-memory

from agentic_memory import MemoryClient, AgentMemory, TemporalKG

mc = MemoryClient()
mc.save("Important context", category="lessons")
results = mc.search("context")
stats = mc.stats()

npm install @agentic-memory/sdk

import { MemoryClient } from '@agentic-memory/sdk';
const client = new MemoryClient();
await client.add('Important context');
const results = await client.search('context');

agentic-memory api --port 9878

curl -X POST http://localhost:9878/api/v1/memories \
  -H "Content-Type: application/json" \
  -d '{"content": "Important context"}'

memory_search         memory_save           memory_delete
memory_recall         memory_note           memory_learn
memory_audit          memory_organize       memory_share
memory_graph          memory_profile        memory_session_start
memory_advanced       memory_review_beliefs memory_curate_autosave
memory_health_check   memory_system_health

{
  "agentic-memory": {
    "command": "agentic-memory-server",
    "env": {
      "MEMORY_KNOWLEDGE_GRAPH": "1",
      "MEMORY_DB_PATH": "./memory.db"
    }
  }
}

from agentic_memory.integrations.langchain.tool import search_tool, save_tool
agent = create_react_agent(llm, tools=[search_tool, save_tool])

from agentic_memory.integrations.crewai.tool import AgenticMemorySearchTool
agent = Agent(..., tools=[AgenticMemorySearchTool()])

mc.okf_export("~/ObsidianVault/agent-memory")

pip install agentic-memory              # Core
pip install agentic-memory[embeddings]  # + semantic search
pip install agentic-memory[reranker]    # + cross-encoder
pip install agentic-memory[langchain]   # + LangChain
pip install agentic-memory[crewai]      # + CrewAI
pip install agentic-memory[all]         # Everything