Updated

Letta alternatives

For agent memory, the alternatives to Letta that makers' products we track chose most are Mem0 (2), Supermemory (2) and Graphiti (1), against 3 for Letta itself. 7 of the 9 are open source or self-hostable (Mem0, Supermemory, Graphiti and 4 more) and 7 have a free tier. Letta also answers AI SDK & agent framework: those alternatives are below too.

7 open source or self-hostable7 with a free tier0 ready for AI agents

Not a ranking: each option is described by the situation it fits, with its trade-off.

Letta alternatives compared for agent memory

How many makers' products we track use each for agent memory.

OptionKindMakers' productsFree tierOpen source
LettaStateful agent platform3YesYes
Mem0Hosted memory API2YesYes
SupermemoryHosted memory API2YesYes
GraphitiOpen-source library1YesYes
ZepHosted memory API0YesNo
CogneeOpen-source library0YesYes
Basic MemoryMarkdown knowledge base0NoYes
LangMemOpen-source library0YesYes
MemaryOpen-source library0YesYes
HyperspellHosted memory API0—No

Letta alternatives compared for AI SDK & agent framework

How many makers' products we track use each for AI SDK & agent framework.

OptionKindMakers' productsFree tierOpen source
LettaBroad framework0YesYes
LangChainBroad framework66YesYes
MastraTypeScript framework24YesYes
LlamaIndexBroad framework16YesYes
AI SDKThin SDK5YesYes
OpenAI Agents SDKThin SDK1YesYes
LangGraphMulti-agent orchestration1YesYes
Pydantic AIThin SDK0YesYes
Claude Agent SDKAgent harness0NoYes
CrewAIMulti-agent orchestration3YesYes
CopilotKitAgent frontend3YesYes
Agent Development KitBroad framework0YesYes
Strands AgentsThin SDK0YesYes
Microsoft Agent FrameworkBroad framework0YesYes
DifyVisual builder0YesSelf-hostable
LangflowVisual builder0YesYes
smolagentsThin SDK0YesYes
DSPyBroad framework0YesYes
AgnoBroad framework0YesYes
OGXBroad framework0YesYes
Semantic KernelBroad framework0YesYes
InstructorThin SDK0YesYes
HaystackBroad framework0YesYes
AutoGenMulti-agent orchestration0—Yes
OpenAI SwarmMulti-agent orchestration0—Yes

When each fits: agent memory

The whole decision →

Hosted memory API, an open-source library you run, or a full stateful-agent platform. A hosted API extracts facts from conversations and returns the relevant ones with one call, with nothing to operate but your users' data in its cloud. An open-source library (often backed by a knowledge graph) runs next to your own database, free but yours to tune and host. A stateful-agent platform makes memory part of the agent itself, which saves glue code but means building your agent on that platform.

Letta (Stateful agent platform): Building the agent itself on a platform where long-term memory, self-editing context and tools are built in. Trade-off: Memory comes with Letta's agent model and API, so moving to another framework later means rebuilding; the company now leads with its own coding agent.

Mem0

Hosted memory API
Open sourceFree tierFrom $19/mo

Best forA drop-in memory layer that extracts facts from conversations and retrieves them later, as a hosted API or an Apache-licensed library.

Trade-offHosted plans cap add and retrieval requests per month, and graph memory starts on the $249/month Pro plan.

Used by 2 makers' products for this

Supermemory

Hosted memory API
Open sourceFree tierFrom $19/mo

Best forA memory API that learns from users' chats, files and URLs and injects what it knows into the model's context, also available as plugins and an MCP server.

Trade-offCredit-based billing makes heavy ingestion harder to predict; self-hosting is only on the Scale plan and above.

Used by 2 makers' products for this

Graphiti

Open-source library
Open sourceFree tier

Best forA Python library for temporal knowledge graphs that agents update in real time, on Neo4j, FalkorDB, Kuzu or Neptune.

Trade-offA library, not a service — you run the graph database and pay for the LLM calls that extract entities and facts.

Used by 1 maker's product for this

Zep

Hosted memory API
Free tierFrom $125/mo

Best forA context graph per user, account or domain built from business data, documents and conversations, with dated facts and access policies.

Trade-offPaid self-serve plans start at $125/month, and it now targets enterprise data more than simple chat memory; VPC deployment is Enterprise-only.

No maker product tracked for this yet

Cognee

Open-source library
Open sourceFree tierFrom $1 per 1M tokens

Best forAgent memory that combines graph, vector and relational retrieval, run locally, self-hosted or on Cognee Cloud.

Trade-offBuilding a graph from your data costs LLM tokens on every ingest, and tuning the pipeline takes more work than a hosted API.

No maker product tracked for this yet

Basic Memory

Markdown knowledge base
Open sourceFrom $15/seat/mo

Best forA knowledge base of plain Markdown files with semantic and graph search, shared with your AI tools through MCP.

Trade-offAGPL-3.0 if you self-host and modify it; built for people and their assistants, not as a multi-tenant memory API for an app.

No maker product tracked for this yet

LangMem

Open-source library
Open sourceFree tier

Best forAdding long-term memory to a LangGraph agent, with memory tools the agent calls mid-conversation and a background manager that extracts and consolidates facts.

Trade-offPython only and tied most closely to LangGraph; no release since October 2025, only dependency upkeep.

No maker product tracked for this yet

Memary

Open-source library
Open sourceFree tier

Best forExperimenting with knowledge-graph memory for an agent, stored in Neo4j or FalkorDB.

Trade-offUnmaintained since 2024, so expect to fix breakages with newer model and graph libraries yourself.

No maker product tracked for this yet

Hyperspell

Hosted memory API

Best forAgents that needed memory and context drawn from their users' connected data sources.

Trade-offPivoted to a "company brain" for a business's own agents; no longer a memory API for your product's users.

No maker product tracked for this yet

When each fits: AI SDK & agent framework

The whole decision →

Broad framework or thin SDK. Broad frameworks come with integrations for loaders, vector stores and memory, but add layers of abstraction you have to learn and debug. Thin SDKs give you a typed agent loop close to the raw API and leave the rest to you. An agent harness goes further the other way: the whole loop — file access, shell, context management — comes ready-made, for agents that work inside a workspace. A visual builder sits outside your code entirely: you assemble the agent, its tools and its retrieval on a canvas and call the result over an API, trading control and version history in your repo for speed and edits non-developers can make.

Letta (Broad framework): Building stateful agents on a server and API where long-term memory and self-editing context are built in. Trade-off: Agents live in Letta's server model, so leaving means rebuilding; the company now leads with its own coding agent.

LangChain

Broad framework
Open sourceFree tierFrom $39/seat/mo (LangSmith)llms.txt

Best forThe largest set of ready-made integrations for models, tools and data sources, in Python and JavaScript.

Trade-offLayers of abstraction make debugging harder, and the API has changed often.

Used by 66 makers' products for this

Mastra

TypeScript framework
Open sourceFree tierFrom $250/mollms.txtCLI

Best forTypeScript and Next.js apps that want agents, workflows, memory and RAG without a Python service.

Trade-offYounger ecosystem, and some enterprise features sit under a separate source-available license.

Used by 24 makers' products for this

LlamaIndex

Broad framework
Open sourceFree tierFrom $50/mo (LlamaCloud)llms.txt

Best forAgents that mostly retrieve and reason over your own documents.

Trade-offLess natural for agents whose work is mostly calling external tools rather than reading data.

Used by 16 makers' products for this

AI SDK

Thin SDK
Open sourceFree tier

Best forTypeScript agents with tool calling, multi-step loops and streaming to the UI, on any model provider.

Trade-offMemory, workflows and evals are left to you or other libraries.

Used by 5 makers' products for this

Open sourceFree tier

Best forA small agent loop with tools, handoffs, guardrails and tracing built in, from the OpenAI team.

Trade-offTracing and hosted tools fit OpenAI best; other providers work but get less attention.

Used by 1 maker's product for this

LangGraph

Multi-agent orchestration
Open sourceFree tierFrom $39/seat/mo (LangSmith Plus, includes one small deployment)llms.txtCLI

Best forStateful agents built as an explicit graph of steps, with checkpoints, retries, streaming and pauses for human approval, in Python or JavaScript.

Trade-offLower-level than a prebuilt agent — you design the graph and state yourself; managed deployment and tracing sit in paid LangSmith plans.

Used by 1 maker's product for this

Pydantic AI

Thin SDK
Open sourceFree tierllms.txt

Best forPython agents whose outputs must validate against typed schemas, across any model provider.

Trade-offPython only, with fewer prebuilt integrations than the broad frameworks.

No maker product tracked for this yet

Claude Agent SDK

Agent harness
Open sourceFrom Pay per tokenllms.txt

Best forAgents that read and edit files, run shell commands and manage long context, built on the same harness as Claude Code.

Trade-offBuilt for Claude models, and it needs a filesystem and process to run in, which serverless functions don't give you.

No maker product tracked for this yet

CrewAI

Multi-agent orchestration
Open sourceFree tierFrom Contact salesllms.txt

Best forPython projects that split work across several role-based agents, like researcher, writer and reviewer.

Trade-offMulti-agent setups cost more tokens and are harder to predict than one well-prompted agent.

Used by 3 makers' products for this

CopilotKit

Agent frontend
Open sourceFree tierFrom $39/mo

Best forPutting an agent inside your app's UI — chat, generative UI, shared state with the page — connected to any agent backend over the AG-UI protocol.

Trade-offIt is the frontend half only; the hosted runtime's free plan keeps threads for 3 days, and longer retention is paid.

Used by 3 makers' products for this

Agent Development Kit

Broad framework
Open sourceFree tierllms.txt

Best forMulti-agent systems with built-in evaluation and a dev UI, in Python, TypeScript, Java or Go, deployable to Google Cloud.

Trade-offWorks with other models but fits Gemini and Google Cloud best; more concepts to learn than a thin SDK.

No maker product tracked for this yet

Open sourceFree tierllms.txtCLI

Best forA model-driven agent loop from AWS in Python or TypeScript, where you give the model tools and let it plan, on Bedrock or other providers.

Trade-offFits AWS best; fewer prebuilt integrations and a smaller community than the broad frameworks.

No maker product tracked for this yet

Open sourceFree tierMCP

Best forAgents and multi-agent workflows in Python or .NET, from Microsoft, replacing AutoGen and Semantic Kernel.

Trade-offFits Azure and Microsoft's stack best; AutoGen itself is in maintenance mode, so new projects should start here rather than on AutoGen.

No maker product tracked for this yet

Dify

Visual builder
Self-hostableFree tierllms.txt

Best forAgents, RAG chatbots and multi-step LLM workflows built on a visual canvas and published as a web app, API or MCP server.

Trade-offThe logic lives in Dify's app definitions rather than your codebase, self-hosting runs a stack of more than a dozen containers, and the modified Apache 2.0 license needs a commercial license for multi-tenant use.

No maker product tracked for this yet

Langflow

Visual builder
Open sourceFree tierllms.txt

Best forAgents, RAG flows and MCP servers assembled from components on a visual canvas and served as an API, an MCP server or an embedded chat widget, under the MIT license.

Trade-offFlows live as JSON in Langflow's database rather than in your codebase, and it runs component code with full host access and no isolation between users, so shared deployments need their own sandboxing.

No maker product tracked for this yet

smolagents

Thin SDK
Open sourceFree tier

Best forA minimal Python agent library from Hugging Face whose main agent writes its actions as Python code, run in a sandbox such as E2B, Modal or Docker.

Trade-offCode actions need a sandbox to be safe, and releases have slowed (last on PyPI in May 2026).

No maker product tracked for this yet

DSPy

Broad framework
Open sourceFree tier

Best forWriting LLM programs as typed modules and letting optimizers such as GEPA tune the prompts or weights against a metric you define.

Trade-offPython only, with its own concepts (signatures, modules, optimizers) to learn; optimization runs cost tokens.

No maker product tracked for this yet

Agno

Broad framework
Open sourceFree tierFrom $150/mo

Best forA Python framework and runtime (AgentOS) for agents, teams and workflows that you run in your own cloud, with a paid control plane for traces, memory, evals and approvals.

Trade-offThe framework is free, but the control plane for a live deployment starts at $150/month.

No maker product tracked for this yet

OGX

Broad framework
Open sourceFree tier

Best forRunning one open-source server that speaks the OpenAI, Anthropic and Google APIs over any model provider, with vector stores, moderation, tool calling and an agent loop built in.

Trade-offA server to run and configure rather than a library you import; renamed from Llama Stack in 2026, so older guides use the old name and packages.

No maker product tracked for this yet

Semantic Kernel

Broad framework
Open sourceFree tier

Best forBuilding agents in C#/.NET, Python or Java with Microsoft's SDK and its Azure integrations.

Trade-offMicrosoft directs new projects to Microsoft Agent Framework; Semantic Kernel gets critical fixes but few new features.

No maker product tracked for this yet

Instructor

Thin SDK
Open sourceFree tier

Best forGetting validated structured output from any major LLM by declaring a Pydantic model, with retries on validation errors.

Trade-offStructured output only, with no agent loop, memory or tool orchestration; the TypeScript port has not been released since January 2025.

No maker product tracked for this yet

Haystack

Broad framework
Open sourceFree tier

Best forBuilding RAG pipelines and agents in Python from explicit, swappable components, with many vector-store and model integrations.

Trade-offPython only; deepset's hosted platform is aimed at enterprises, so at small scale you host the pipeline yourself.

No maker product tracked for this yet

AutoGen

Multi-agent orchestration
Open source

Best forResearch-style multi-agent systems in Python where agents converse with each other and with humans.

Trade-offArchived by Microsoft with no bug or security fixes; new projects should use Microsoft Agent Framework.

No maker product tracked for this yet

OpenAI Swarm

Multi-agent orchestration
Open source

Best forLearning how agent handoffs and routines work, in a tiny Python library.

Trade-offReplaced by the OpenAI Agents SDK, which OpenAI recommends migrating to.

No maker product tracked for this yet

Questions

What is the best alternative to Letta?

It depends on what you need it for; we don't rank them. The one makers' products we track chose most for agent memory is Mem0 (2 makers' products). You want to add user memory to an existing agent with a few API calls and a free tier to start: Mem0 and Supermemory. Memory must stay on your own infrastructure, built as a graph over your data: Cognee and Graphiti. Your agent needs dated facts that change over time, from business data as well as chats: Zep and Graphiti.

Is there a free alternative to Letta?

Yes. For agent memory, Mem0, Supermemory, Graphiti, Zep, Cognee, LangMem and 1 more have a free tier a small product can start on.

Is there an open-source alternative to Letta?

Yes. For agent memory: Mem0, Supermemory, Graphiti, Cognee, Basic Memory, LangMem and 1 more.