The memory layer
for AI agents that learn.

from mengram import Mengram

m = Mengram(api_key="om-...")

# Save — auto-extracts facts, events, workflows
m.add([
    {"role": "user", "content": "Fixed OOM with Redis cache"},
    {"role": "assistant", "content": "Got it."},
])

# Unified search — all 3 memory types
results = m.search_all("database issues")
# → {semantic: [...], episodic: [...], procedural: [...]}

# Cognitive Profile — instant personalization
profile = m.get_profile()
# → ready system prompt for any LLM

# Multi-user isolation — one API key, many users
m.add([...], user_id="alice")
m.search_all("prefs", user_id="alice")  # only Alice's data

const { MengramClient } = require('mengram-ai');

const m = new MengramClient('om-...');

// Save — auto-extracts facts, events, workflows
await m.add([
    { role: 'user', content: 'Fixed OOM with Redis cache' },
    { role: 'assistant', content: 'Got it.' },
]);

// Unified search — all 3 memory types
const all = await m.searchAll('database issues');
// → {semantic: [...], episodic: [...], procedural: [...]}

// Multi-user isolation — one API key, many users
await m.add([...], { userId: 'alice' });
await m.searchAll('prefs', { userId: 'alice' }); // only Alice's data

from langchain_mengram import MengramRetriever

retriever = MengramRetriever(
    api_key="om-...",
    user_id="alice",
    top_k=5,
)

# Use in any LangChain chain
docs = retriever.invoke("deployment issues")
# → Documents from semantic + episodic + procedural memory

# Or in an LCEL chain
chain = (
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt | llm | StrOutputParser()
)

from langchain_mengram import MengramChatMessageHistory
from langchain_core.runnables.history import RunnableWithMessageHistory

chain_with_memory = RunnableWithMessageHistory(
    chain,
    lambda sid: MengramChatMessageHistory(
        api_key="om-...",
        session_id=sid,
    ),
    input_messages_key="input",
    history_messages_key="history",
)

from crewai import Agent, Crew, Task
from integrations.crewai_memory import MengramMemory

agent = Agent(
    role="DevOps Engineer",
    goal="Deploy and monitor services",
)

task = Task(
    description="Deploy v2.15 to staging",
    agent=agent,
)

# One line adds persistent memory to your entire crew
crew = Crew(
    agents=[agent],
    tasks=[task],
    memory=MengramMemory(api_key="om-..."),
)
crew.kickoff()
# → Agent recalls past deployments, learns from failures

{
  "plugins": {
    "entries": {
      "openclaw-mengram": {
        "enabled": true,
        "config": {
          "apiKey": "${MENGRAM_API_KEY}"
        }
      }
    },
    "slots": { "memory": "openclaw-mengram" }
  }
}
// Auto-recall: memories injected before every agent turn
// Auto-capture: new info saved after every turn

Header Name: Authorization
Header Value: Bearer om-your-api-key

// Node 1: Search memories
POST /v1/search → {"query": user_message, "user_id": "user-123"}

// Node 2: AI Agent responds with full context
System prompt includes retrieved memories

// Node 3: Save new memories
POST /v1/add → {"messages": [...], "user_id": "user-123"}
// Auto-extracts facts, deduplicates, builds knowledge graph