Alternatives to Letta

Cognee gives AI agents persistent knowledge-graph memory with cited recall across sessions, teams, and tools.

Mem0 is an AI memory layer that gives agents persistent context across sessions, users and apps

Athena Public gives your LLM a persistent, editable memory layer that survives across sessions and IDEs.

EverMemOS gives AI agents persistent, self-evolving memory across sessions, platforms, and teams.

CogniMemo is an API-first persistent memory layer that gives your AI agents structured long-term recall across sessions.

Omnara is a stateful, open-source agent API for deploying durable AI agents on any model, machine, or cloud.

DataGrout is an operating layer for production AI agents: persistent memory, real-time cost control, and cryptographic receipts for every action.

Hebbrix is outcome-weighted persistent memory for AI agents — it recalls what actually worked, not what merely sounds related.

Three markdown files that give any AI coding agent persistent memory between sessions — no framework, no database, just git.

Local-first persistent memory for AI agents with a forgetting curve, open source and MCP-compatible.

Persistent memory for Python AI agents, with versioning, shared memory across agent teams, and loop detection that pauses runaway agents before the token bill

MemOS gives AI agents persistent long-term memory through a cloud API or a fully local runtime.

Self-hosted layer that gives your AI coding agents shared memory, managed MCP tools, and cross-model audit.

TrueMemory gives CLI coding agents persistent, local-first memory recall in about 55ms — one file, zero cloud.

Jumbo.Cli gives AI coding agents persistent project memory from a local-first CLI, so you stop re-onboarding Claude Code every session.

Open-source, self-hosted memory layer that gives AI agents persistent recall via hybrid retrieval.

A persistent, file-based memory layer for coding agents that lives as a brain/ Markdown folder in your repo.

Inkbox gives AI agents a persistent identity with email, phone, iMessage, and an internet address.

Shared memory for Claude Code teams—decisions, handoffs, and context persist across sessions and people.

Agentmemory gives coding agents shared, local memory across Claude Code, Cursor, Codex and 22 other agents.

Hosted MCP and REST memory that keeps AI agent context consistent across ChatGPT, Claude, Codex, and your own agents.

Long-term memory for AI agents: one process_turn call stores each turn and returns the memories your next prompt needs.

Open-source Postgres memory engine for AI agents that sends memory conflicts to a human instead of overwriting them.

Carrier-native AI memory platform: every SIM becomes a persistent agent with its own billable identity.

Open-source MCP memory layer that gives AI agents a graph plus vector store so recall carries relationships, not just similar text.

Hardware-isolated microVMs for AI agents that boot in under 200ms, with snapshots, forkable state, and persistent volumes.

memU gives Codex, Claude, Cursor and other agents one shared, inspectable memory layer with source-linked recall.

Memori is an LLM-agnostic agent memory gateway that turns every execution trace into persistent, governed memory and procedural recall.

Memoryport is a local-first memory layer that gives Claude Code, Cursor, Ollama and OpenAI persistent context across every session.

Zero-LLM persistent agent memory — Hebbian learning and decay curves in a single 30MB Rust binary.