Mem0
Memory layer for AI agents that stores and retrieves user context across sessions, open source with a hosted platform
FREEMIUM Verified JUL 18, 2026
Letta
Platform for stateful AI agents with persistent memory that survives across sessions
Letta builds agents that keep their memory between sessions instead of starting cold each conversation, with a memory filesystem, subagents, and channels into Slack, Telegram, and WhatsApp. The free plan is $0 and caps you at 3 stateful agents using your own API keys; Pro is $20/month for up to 20 agents, Teams Pro $20/seat, and the developer API plan is $20/month plus $0.10 per active agent and $0.00015 per second of tool execution. Apache-2.0 core.
Verified JUL 30, 2026 FREEMIUM Live
What is Letta?
Letta is a platform for building agents that remember. Most LLM applications reconstruct context on every request, which means an assistant either forgets what happened last week or pays to re-read a growing transcript. Letta treats memory as infrastructure instead: agents are stateful, with a memory filesystem called MemFS, consolidation between runs, and persistence that survives restarts. Around that core sit subagents, skills, mods, scheduled runs, secrets management, and channels that place an agent directly into Slack, Telegram, Discord, WhatsApp, or Signal.
The product is available as a web app, a desktop app, a CLI, and a developer API, with execution either in Letta’s cloud sandboxes or on your own machine. Pricing converges on a single $20 figure used three ways: Pro for individuals, per-seat for Teams Pro, and a monthly base for the developer API plan. The developer plan is the one to model carefully, because it removes the agent cap but adds $0.10 per active agent per month and $0.00015 per second of tool execution on top of pass-through LLM token costs. The core is Apache-2.0 with more than 24,000 GitHub stars, and self-hosting is documented for teams that would rather run the server themselves.
Who is it for?
Letta fits developers whose agents need continuity — where the value comes from what the agent has learned across many sessions rather than from a single well-crafted prompt. The free tier is a real evaluation path because it runs on API keys you already hold, but its 3-agent cap is restrictive enough that any serious project moves to Pro quickly.
- Developers building long-lived assistants that track a user’s history, preferences, or ongoing project state without re-sending a transcript every call.
- Teams replacing stateless chatbots where forgetting context between conversations is the main complaint, and memory is the feature being bought.
- Agent engineers running multi-agent systems, using subagents and shared organisation access so several agents cooperate on one workflow.
- Cost-sensitive builders who want to keep model spend on their own provider account through bring-your-own-key rather than paying a platform markup on inference.
How much does Letta cost?
Starting price: $20/mo · Free tier: yes · Model: freemium
Pricing verified JUL 30, 2026
Price history tracked from June 2026
| Plan | Price | Includes |
|---|---|---|
| Free | Free | Limited to 3 stateful agents · Limited usage of Letta Auto · Bring your own API keys and external coding plans · Works with ChatGPT/Codex, Kimi, and Z.AI coding plans |
| Pro | $20/mo | Up to 20 stateful agents · Letta Auto weekly and monthly quota · Agents run in remote sandboxes · Purchasable credits across models, including image generation · Pay-as-you-go overage |
| API Plan | $20/mo | Unlimited agents · $0.10 per active agent per month · $0.00015 per second of tool execution · API key authentication · Pay-as-you-go LLM usage on top |
| Teams Pro | $20/seat/mo | Add teammates to an organisation · Share agents and set access control · Letta Auto weekly and monthly quota · Includes the individual Pro features |
| Enterprise | Custom | Volume-based pricing · Increased quotas · Role-based access control · SAML/OIDC single sign-on · Dedicated support |
What are Letta's key features?
- Stateful agents whose memory persists across sessions by design
- MemFS memory filesystem plus memory consolidation between runs
- Subagents, skills, and mods for composing agent behaviour
- Channels for Slack, Telegram, Discord, WhatsApp, and Signal
- CLI, desktop app, and web app alongside the developer API
- Cloud sandboxes or bring-your-own-machine execution
- Schedules, secrets management, and permission controls
- Apache-2.0 licensed core with 24,000+ GitHub stars
What people use Letta for
- 01 Running an assistant that remembers earlier conversations without re-sending a transcript
- 02 Giving an agent a persistent filesystem it can read and write across sessions
- 03 Deploying an agent into Slack, Telegram, Discord, WhatsApp, or Signal as a channel
- 04 Coordinating subagents that hand work between each other inside one organisation
- 05 Scheduling agents to run on a recurring basis rather than only on user prompts
Pros and cons
| Pros | Cons |
|---|---|
| Persistent memory is the core architecture rather than a bolted-on context window trick | Agent counts are hard-capped on personal plans — 3 on Free and 20 on Pro — so hitting the ceiling forces an upgrade even at low usage |
| The free tier runs on your own API keys, including existing coding-plan subscriptions, so evaluation costs nothing extra | The developer API bill has three moving parts: a $20 monthly fee, $0.10 per active agent, and $0.00015 per second of tool execution, before LLM tokens |
| Apache-2.0 licence on the core with a documented self-hosting path — no licence negotiation required | LLM token costs are passed through separately, so the subscription is not the whole spend |
| One $20 price point covers Pro, the developer API plan, and per-seat Teams, which keeps the ladder simple | Letta Auto allowances are published only as a 'weekly and monthly quota' without absolute numbers on the pricing page |
| SSO and role-based access control are Enterprise-only, with no published price |
What are the best Letta alternatives?
CrewAI
Open-source multi-agent framework for orchestrating collaborative AI agent teams
FREEMIUM Verified JUL 3, 2026
★ 54.9k as of Jul 5, 2026
LangChain
Open-source frameworks for building LLM agents, plus the commercial LangSmith platform for tracing, evaluation, and deployment
FREEMIUM Verified JUL 12, 2026
Dify
Open-source LLM app platform for building agentic workflows, RAG pipelines, and AI agents with a visual no-code builder
FREEMIUM Verified JUN 24, 2026
Langfuse
Open-source LLM engineering platform for tracing, observability, evals, and prompt management — self-host free or run on Langfuse Cloud
OPEN-SOURCE Verified JUL 15, 2026
How people make money with Letta
- Build persistent client-facing assistants that remember account history between conversations, billed as a setup project plus a per-seat monthly support fee
- Offer memory migration and agent design for teams whose chatbots forget context, charging a fixed engagement fee with the developer API plan as the delivery platform
Frequently asked questions
Is Letta free?
Yes, with limits. The free plan costs nothing and lets you run up to 3 stateful agents while bringing your own API keys, including existing coding plans from ChatGPT/Codex, Kimi, or Z.AI. Letta Auto usage is limited. Pro at $20/month raises the cap to 20 agents and adds remote sandboxes.
How much does Letta Pro cost?
Pro is $20 per month. It includes a weekly and monthly Letta Auto quota, agents running in remote sandboxes, up to 20 stateful agents, purchasable credits usable across models including image generation, and pay-as-you-go overage once the included quota is consumed.
What does the Letta developer API cost?
The API plan is $20 per month and removes the agent cap entirely. On top of the base fee you pay $0.10 per active agent per month and $0.00015 per second of tool execution, plus pay-as-you-go charges for the underlying LLM tokens your agents consume.
What makes a Letta agent stateful?
The agent retains memory between sessions rather than being rebuilt from a prompt each time. Letta provides a memory filesystem called MemFS and consolidates memory between runs, so an agent accumulates knowledge about its task and user instead of depending on a transcript being resent.
Can I self-host Letta?
Yes. The core is Apache-2.0 licensed with more than 24,000 GitHub stars and the documentation includes a self-hosting path plus Docker support. Self-hosting avoids the hosted plan fees, though you take on running the server and supplying your own model access.
Does Letta support team collaboration?
Yes, through Teams Pro at $20 per seat per month. It adds members to an organisation, allows agents to be shared with access controls, and lets agents within the organisation collaborate. Role-based access control and SAML or OIDC single sign-on require the custom-priced Enterprise plan.