Real use cases, real code. See how developers use Octopoda to manage, monitor, and give their agents persistent memory.
Your support bot asks customers to repeat themselves every session. "What's your order number?" for the 5th time. Customers hate it.
With Octopoda, the agent remembers every customer interaction. Their name, order history, previous issues, preferences — all recalled instantly.
from octopoda import LangChainMemoryfrom langchain.chains import ConversationChainfrom langchain_openai import ChatOpenAI # Each customer gets persistent memorymemory = LangChainMemory(f"support_{customer_id}")chain = ConversationChain(llm=ChatOpenAI(), memory=memory) # First interactionchain.invoke({"input": "Hi, I'm Alice. Order #12345 hasn't arrived."}) # Days later — agent remembers everythingchain.invoke({"input": "Any update on my order?"})# Agent recalls: Alice, order #12345, shipping issue# No need to ask againYou have a CrewAI setup with a researcher, analyst, and writer. But Agent A discovers something and Agent B has no idea. They duplicate work, miss connections, and produce inconsistent reports.
Octopoda's shared memory bus lets all agents access the same knowledge. The researcher stores findings, the analyst reads them, and the writer pulls from both.
from octopoda import CrewAIMemory crew_memory = CrewAIMemory("research_team") # Researcher stores findingscrew_memory.save_agent_output("researcher", "market_data", "AI agent market growing at 34% CAGR, expected $12B by 2028") # Analyst accesses researcher's workcrew_memory.save_agent_output("analyst", "competitive_landscape", "Top 5 players hold 67% market share. Biggest gap: agent memory.") # Writer searches across ALL agent knowledgeresults = crew_memory.search("market growth projections")# Finds both researcher and analyst outputs instantly # Track what the crew knowssummary = crew_memory.get_crew_summary()# {"task_results": 3, "agent_outputs": 8, "shared_knowledge": 5}Every AI assistant starts as a stranger. It doesn't know your preferences, your schedule patterns, your communication style, or what you're working on.
With Octopoda, your assistant builds a growing understanding of you over time. It remembers your preferences, learns your patterns, and gets better every conversation.
from octopoda import OpenAIAgentsMemoryfrom agents import Agent, Runner memory = OpenAIAgentsMemory() assistant = Agent( name="Personal Assistant", instructions="""You are a personal AI assistant. Use your memory tools to learn about the user over time. Always check your memory before asking questions they've already answered.""", tools=memory.tools(),) # Week 1Runner.run_sync(assistant, "I'm Alice, CTO at TechCorp. I prefer bullet points.")Runner.run_sync(assistant, "Schedule a meeting with the ML team for Thursday") # Week 4 — assistant knows everythingRunner.run_sync(assistant, "Prepare talking points for my next meeting")# Agent searches memory → knows Alice is CTO, prefers bullets,# has ML team meetings on Thursdays, and what was discussed last timeYou deploy 20 agents to production. One crashes. Another starts giving weird answers. A third has been idle for 2 hours. You have no visibility into any of this.
Octopoda's dashboard shows every agent's health, memory state, and activity in real-time. Set alerts for crashes, anomalies, and limits approaching.
Start with any use case. Memory, monitoring, crash recovery, and diagnostics — all included.