MemoSift
Integrations

Anyframework.
Oneline.

Three integration levels — from tool interception to full context lifecycle management. Pick the depth that fits your stack.

LEVEL 1

Tool Interception

Intercept tool results. Classify, scan, store. Agent sees summaries.

LEVEL 2

Lifecycle Hooks

Auto-track turns in background. Context injection hints.

LEVEL 3

Full Orchestration

Auto-recall, auto-compress, auto-track. Complete context lifecycle.

OpenAI Agents SDK

OpenAI

Tool Interception

Decorator wraps your tools. One line per function.

ms.openai_agents_tool_wrapper()

Tool interception
Content classification
Security scanning
Artifact storage
Auto memory tracking
Auto recall injection
openai_agent.py
from memosift import MemoSift

ms = MemoSift(api_key="msk_...")
wrap = ms.openai_agents_tool_wrapper(session_id="my-session")

@wrap
async def read_file(path: str) -> str:
    with open(path) as f:
        return f.read()

# Tool results auto-intercepted.
# Large files → stored as artifacts, agent sees summary.

LangGraph

LangChain

Tool Interception

awrap_tool_call for ToolNode. Session ID from thread_id.

ms.langgraph_wrapper()

Tool interception
Content classification
Security scanning
Artifact storage
Auto memory tracking
Auto recall injection
langgraph_agent.py
from langgraph.prebuilt import ToolNode
from memosift import MemoSift

ms = MemoSift(api_key="msk_...")
awrap = ms.langgraph_wrapper()

# session_id auto-extracted from config.configurable.thread_id
tool_node = ToolNode(tools=[...], awrap_tool_call=awrap)

Claude Agent SDK

Anthropic

Lifecycle Hooks

PostToolUse hook. Background tracking with context hints.

ms.claude_agent_hooks()

Tool interception
Content classification
Security scanning
Artifact storage
Auto memory tracking
Auto recall injection
claude_agent.py
from memosift import MemoSift

ms = MemoSift(api_key="msk_...")
hooks = ms.claude_agent_hooks()

# Pass to your Claude Agent SDK setup.
# Large tool results → artifact stored, agent gets:
# "[MemoSift] Stored as artifact art_abc123.
#  Use memosift_fetch('art_abc123') to retrieve."

Client Wrapper

Any Python

Full Orchestration

Full orchestration. Recall, compress, track — automatic.

ms.wrap()

Tool interception
Content classification
Security scanning
Artifact storage
Auto memory tracking
Auto recall injection
wrapped_client.py
from openai import AsyncOpenAI
from memosift import MemoSift

ms = MemoSift(api_key="msk_...")
client = ms.wrap(AsyncOpenAI(), session_id="my-session")

# Use like normal — MemoSift handles everything:
# Pre-call:  recall memories → inject context
# Tool use:  intercept large results → artifact stubs
# Post-call: track turn in background
# Pressure:  auto-compress at 80% context budget
response = await client.chat.completions.create(
    model="gpt-4.1",
    messages=[{"role": "user", "content": "..."}],
)

Claude Code

MCP Server

Full Orchestration

4 MCP tools over stdio. Cross-session memory for Claude Code.

MCP Server (stdio)

Tool interception
Content classification
Security scanning
Artifact storage
Auto memory tracking
Auto recall injection
terminal
# One-command install for Claude Code
memosift login
memosift install-claude-code

# Or run the MCP server directly (stdio transport)
memosift serve

# Exposes 4 tools to Claude Code:
# memosift_recall   → search, intent recall, time-travel
# memosift_explore  → walk entity graph connections
# memosift_fetch    → download full artifact content
# memosift_compress → assemble session context block
Agent Tools

4 tools your agent calls directly.

Returned as OpenAI-compatible function schemas via session.tools(). Works with any framework that supports function calling.

memosift_recall

Search memories by query, recall by intent epoch, or snapshot at any turn.

query · intent_version · turn

memosift_explore

Walk entity connections from a known result. No search query needed.

entity · artifact · temporal · intent axis

memosift_fetch

Download full artifact content with metadata. Inline for <500KB.

artifact_id → content + metadata

memosift_compress

Assemble structured session context. No LLM call, pure SQL + template.

intent-organized knowledge block

Start with one line. Scale from there.

pip install memosift