Octopoda-OS

Overview

Octopoda serves as a memory operating system for AI agents, offering features like persistent memory that survives restarts, loop detection to prevent infinite token consumption, and a comprehensive audit trail with hash-chained events for verification. It provides both local and cloud deployment options with the same Python API. The MCP server implementation exposes 29 tools that enable developers to enhance their AI agents with these capabilities through the Model Context Protocol.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• rememberremember(key, value)

  Store a value with a given key in the agent's persistent memory

  Try:Remember that the user prefers dark modeTry:Store the project deadline as December 31st
• recallrecall(key)

  Retrieve a value from the agent's memory using a specific key

  Try:What was the user's preference for theme?Try:What's the project deadline?
• recall_similarrecall_similar(query)

  Find memories based on semantic similarity rather than exact key matching

  Try:Find memories about user dietary preferencesTry:What does the user think about the new feature?
• forgetforget(key)

  Delete a specific memory using its key

  Try:Remove the outdated configuration settingsTry:Forget the temporary API token
• log_decisionlog_decision(decision, reasoning, context)

  Log a decision made by the agent with reasoning and context for audit trail

  Try:Log the decision to use PostgreSQL instead of MongoDBTry:Document why we chose to retry the failed request
• send_messagesend_message(recipient, message, message_type)

  Send a message to another agent through their inbox

  Try:Send an alert to the support agent about the system issueTry:Message the research agent with the new findings
• read_messagesread_messages(unread_only)

  Read messages sent to this agent, optionally filtering for unread messages only

  Try:Check for any new messages from other agentsTry:Show me all unread alerts
• set_goalset_goal(goal, milestones)

  Set a goal with milestones for the agent to track

  Try:Set up the goal to migrate the entire system to microservicesTry:Create a goal with milestones for the product launch
• update_progressupdate_progress(milestone_index, note)

  Update progress on a specific milestone with a note

  Try:Mark the first milestone as completed with notes about the work doneTry:Update progress on the testing phase
• snapshotsnapshot(description)

  Create a snapshot of the agent's current state for potential recovery

  Try:Take a snapshot before making major changesTry:Snapshot the state before deployment
• restorerestore(snapshot_id)

  Restore the agent's state from a previous snapshot

  Try:Restore from the snapshot taken before the system crashTry:Go back to the state from last week
• forget_staleforget_stale(max_age_seconds)

  Remove memories older than the specified age

  Try:Clean up memories older than 30 daysTry:Remove all temporary data older than a week

Comparable tools

Installation

pip install octopoda

{
  "mcpServers": {
    "octopoda": {
      "command": "python",
      "args": ["-m", "octopoda.mcp"]
    }
  }
}

FAQ

How does Octopoda detect loops in AI agents?

Octopoda uses a 5-signal engine that detects retry, oscillation, ping-pong, reflection, and recall patterns in seconds, automatically surfacing which calls caused the loop.

Can I use Octopoda with existing AI agent frameworks?

Yes, Octopoda integrates with LangChain, CrewAI, AutoGen, OpenAI Agents SDK, and other frameworks without requiring changes to your agent implementation.

On Hacker News

Recent discussion from the developer community.

• Octopoda open source agent OS with memory, loop detection, and audit trails▲ 5
  Story by Josephjackjrob1 · 2026-04-07

Compare Octopoda-OS with