AI Framework Integrations

Use Awareness as persistent memory for multi-agent frameworks. All integrations are available through the Python SDK and share a unified adapter pattern.

Supported Frameworks

Framework	Install	Adapter	Key Feature
LangChain	`pip install awareness-memory-cloud[langchain]`	`MemoryCloudLangChain`	Retriever, RAG chains, tool registration
CrewAI	`pip install awareness-memory-cloud[crewai]`	`MemoryCloudCrewAI`	Cross-task memory, crew-member scoping
PraisonAI	`pip install awareness-memory-cloud[praisonai]`	`MemoryCloudPraisonAI`	Tool-based memory in agent workflows
AutoGen / AG2	`pip install awareness-memory-cloud[autogen]`	`MemoryCloudAutoGen`	Multi-agent conversations with memory

Unified Adapter Pattern

All framework integrations extend MemoryCloudBaseAdapter and provide:

wrap_llm() — intercept LLM calls to auto-recall and auto-capture (Interceptor pattern)
build_tools() — generate framework-native tool definitions
get_tool_functions() — get raw tool functions for manual registration
inject_into_messages() — manual message-level injection
awareness_recall() / awareness_record() — explicit memory read/write

Interceptor: The Fastest Path

All framework adapters support the Interceptor pattern — wrap your LLM client once and get automatic memory injection:

from memory_cloud import MemoryCloudClient
from memory_cloud.integrations.langchain import MemoryCloudLangChain
import openai

client = MemoryCloudClient(base_url="...", api_key="...")
mc = MemoryCloudLangChain(client=client, memory_id="memory_123")

# One line: all LLM calls get memory context
mc.wrap_llm(openai.OpenAI())

This works identically across all four frameworks — just swap the adapter class.

OpenAI / Anthropic Direct (No Framework)

For direct API usage without a framework, use the SDK's built-in AwarenessInterceptor:

from memory_cloud import MemoryCloudClient, AwarenessInterceptor
import openai

client = MemoryCloudClient(base_url="...", api_key="...")
interceptor = AwarenessInterceptor(client=client, memory_id="mem-xxx")

oai = openai.OpenAI()
interceptor.wrap_openai(oai)
# All oai.chat.completions.create() calls now have memory injection

See the SDK Usage Guide for full interceptor options and query rewrite modes.

Multi-User / Multi-Role

All framework integrations support user_id and agent_role for team scenarios:

mc_dev = MemoryCloudLangChain(
    client=client, memory_id="memory_123",
    default_metadata={"agent_role": "developer", "user_id": "alice"}
)

mc_reviewer = MemoryCloudLangChain(
    client=client, memory_id="memory_123",
    default_metadata={"agent_role": "reviewer", "user_id": "bob"}
)

When querying with a user_id, results include both that user's data AND public (user_id=NULL) data.

TypeScript

The TypeScript SDK (@awareness-sdk/memory-cloud) provides the same core API and interceptor but does not yet include framework-specific adapters. Use the direct client or interceptor:

import { MemoryCloudClient, AwarenessInterceptor } from "@awareness-sdk/memory-cloud";
import OpenAI from "openai";

const client = new MemoryCloudClient({ baseUrl: "...", apiKey: "..." });
const interceptor = await AwarenessInterceptor.create({ client, memoryId: "mem-xxx" });

const oai = new OpenAI();
interceptor.wrapOpenAI(oai);