Persistent memory for AI agents that survives across sessions. Store, recall, and semantically search memories. Includes loop detection that catches agents stuck repeating themselves, full audit trails, shared memory between agents, and a real-time dashboard. Works with Claude Code, Claude Desktop, and Cursor. Open source, free tier available.
Overview
Octopoda Memory
Persistent memory for AI agents. Store, recall, and search memories that survive across sessions.
Features
- Persistent Memory — agent.remember() and agent.recall() that survives restarts, crashes, deployments
- Semantic Search — find memories by meaning, not just exact keys
- Loop Detection — catches agents stuck repeating themselves before they burn tokens
- Audit Trail — every decision logged with reasoning
- Shared Memory — agents share knowledge through memory spaces
- Real-time Dashboard — monitor all agents at localhost:7842
Setup
- Install:
pip install octopoda - Get free API key at https://octopodas.com
- Add to Claude Code:
claude mcp add octopoda -s user -e OCTOPODA_API_KEY=your-key -- python -m synrix_runtime.api.mcp_server
13 Tools Available
- octopoda_remember — store a memory
- octopoda_recall — get a memory by key
- octopoda_recall_similar — semantic search
- octopoda_search — prefix search
- octopoda_recall_history — version history
- octopoda_snapshot / octopoda_restore — checkpoint and rollback
- octopoda_share / octopoda_read_shared — shared memory between agents
- octopoda_list_agents — list all agents
- octopoda_agent_stats — performance stats
- octopoda_log_decision — audit trail
- octopoda_loop_status — loop detection
Links
Server Config
{
"mcpServers": {
"octopoda": {
"command": "python",
"args": [
"-m",
"synrix_runtime.api.mcp_server"
],
"env": {
"OCTOPODA_API_KEY": "your-key-from-octopodas.com"
}
}
}
}