Integrations
Connect StremAI persistent memory to any AI agent, framework, or platform. Every integration is real, production code — not stubs.
OAuth 2.1 over MCP HTTP is available for all MCP clients listed below. Point your client at https://www.aiagentsbay.com/api/mcp with no API key. The client performs Dynamic Client Registration automatically, opens a browser to StremAI for consent, and receives a short-lived token. The grant covers your whole account brain and can be revoked from your dashboard.
Today, Claude Code and Claude Desktop are verified end to end. Cursor, Codex CLI, Windsurf, and OpenClaw are in verification.
See /oauth for the full flow.
pip install agentbay # All integrations included
IDE & Agent Tools
Claude Code
Lifecycle HooksGlobal MCP setup + lifecycle hooks (session start, stop, pre-compaction, task completion). Auto-recall and auto-store.
bash packages/claude-code-plugin/install.shOpenClaw
Native PluginWebsite installer + native plugin. Strict memory mode by default: installs/enables StremAI, disables memory-core, adds runtime context packs before turns, auto-capture after, explicit checkpoints, policy status, and stronger defaults.
curl -sSL https://www.aiagentsbay.com/install-openclaw.sh | bashCursor / Windsurf
MCPMCP server config. HTTP or NPX connection. Same MCP tool set as Claude Code (tool count varies by transport and package version).
Add MCP config to settingsHermes Agent
MCP + protocolHTTP or stdio MCP config. Optional memory protocol / skill template for structured recall and store.
Add MCP config + memory protocolAgent Frameworks
LangChain
BaseMemory backend + agent Tool. load_memory_variables() and save_context().
from agentbay.integrations.langchain import AgentBayMemoryCrewAI
Drop-in memory backend. save(), search(), reset(). Auto-type detection.
from agentbay.integrations.crewai import AgentBayMemoryAutoGen
Hook-based integration. before_reply() injects context, after_reply() stores learnings.
from agentbay.integrations.autogen import AgentBayMemoryLlamaIndex
BaseMemory class. get(), put(), get_all(), reset().
from agentbay.integrations.llamaindex import AgentBayMemoryLangGraph
BaseCheckpointSaver for state persistence. Thread-scoped checkpoints.
from agentbay.integrations.langgraph import AgentBayCheckpointerOpenAI Agents SDK
NewTool definitions for search + store. Context injection helper.
from agentbay.integrations.openai_agents import AgentBayMemoryToolGoogle ADK
NewBefore/after model callbacks + tool. Auto-recall and auto-store.
from agentbay.integrations.google_adk import AgentBayMemoryCamel AI
Multi-agent conversation memory with user_id scoping.
from agentbay.integrations.camel_ai import AgentBayMemoryAgno
Autonomous agent memory. recall() and store().
from agentbay.integrations.agno import AgentBayMemoryLow-Code Platforms
Dify
NewExternal Knowledge Base API adapter. Connect via Dify's Knowledge settings.
from agentbay.integrations.dify import DifyMemoryAPIFlowise
NewCustom Tool node config + JavaScript snippet generator.
from agentbay.integrations.flowise import FlowiseMemoryToolVercel AI SDK
Memory provider for Next.js AI apps. get_context() and save_context().
from agentbay.integrations.vercel_ai import AgentBayProviderMastra
TypeScript agent framework memory backend.
from agentbay.integrations.mastra import AgentBayMemoryProviderRaycast
NewScript Command generator for quick memory search/store from Raycast.
from agentbay.integrations.raycast import RaycastMemoryCloud & Infrastructure
AWS Bedrock
NewLambda action group handler + OpenAPI schema for Bedrock agents. Context injection helper.
from agentbay.integrations.aws_bedrock import AgentBayBedrockMemoryTavily Search
NewMemory-cached web search. Saves API credits on repeated queries. search_and_remember() for permanent research.
from agentbay.integrations.tavily import MemoryEnhancedSearchVoice & Specialized
LiveKit
NewReal-time voice/video agent memory. before_llm/after_llm callbacks for VoicePipelineAgent.
from agentbay.integrations.livekit import AgentBayLiveKitMemoryElevenLabs
Voice agent memory. on_message() recalls, on_response() stores.
from agentbay.integrations.elevenlabs import AgentBayVoiceMemoryPipecat
Pipeline processor. Handles TranscriptionFrame and TextFrame.
from agentbay.integrations.pipecat import AgentBayProcessorOpenAI Codex
MCP + memory protocol. One-command install: curl -sSL aiagentsbay.com/install-codex.sh | bash
codex mcp add agentbay --url https://www.aiagentsbay.com/api/mcp --bearer-token-env-var AGENTBAY_API_KEYAgentOps
Observability wrapper. Track memory ops with timing and success/failure.
from agentbay.integrations.agentops import track_memory_opsKeywords AI
NewLLM monitoring integration. Tracks store/recall latency and success rates.
from agentbay.integrations.keywords_ai import track_with_keywords_aiDon't see your framework?
StremAI works with any tool that can make HTTP calls. Use the REST API or MCP endpoint directly.