Letta
Self-improving AI agents with persistent memory and continuous learning
If you need AI agents that truly learn and remember without context loss, Letta's research-backed approach is leading. But it's early infrastructure—expect engineering investment. Not for simple chatbot use cases.
Verified 17d ago · liveness 95/100 · cite: rightaichoice.com/tools/letta
- Building long-running AI agents that need persistent memory and learning
- Researchers and developers experimenting with continual learning for AI
- Customer support bots that remember every interaction and improve over time
- Personal AI assistants that adapt to user behavior and preferences
- Simple one-shot chatbot use cases
- Teams needing a plug-and-play agent with no engineering overhead
- Production deployments requiring mature ecosystem and extensive integrations
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Letta if you need a simple, plug-and-play chatbot with no engineering overhead.
Pro tier has limited Letta Auto quota; heavy usage requires pay-as-you-go credits.
Letta's pricing is developer-friendly with a generous free tier for local use. The Pro tier at $20/mo is competitive for personal AI agents, but heavy users (especially coders) will face additional costs for Letta Auto or external coding plans. Compared to tools like LangChain (free) or OpenAI Assistants (per-token), Letta's managed state and memory features justify the cost for serious agent builders.
In short
Letta — Self-improving AI agents with persistent memory and continuous learning. Best for Building long-running AI agents that need persistent memory and learning, Researchers and developers experimenting with continual learning for AI, Customer support bots that remember every interaction and improve over time. Free to start; paid plans from $20/mo.
What's new in Letta
Checked 18 days agoAcross the latest 10 updates: 6 feature updates, 3 launches and 1 news mention.
Introducing Mods: Enabling Agents to Self-Improve through Harness-Level Adaptation
Letta introduces Mods, allowing agents to self-improve via harness-level adaptation.
Introducing the Letta Code App
Launch of Letta Code app for locally-run, personalized agents with persistent memory.
Remote Environments for Letta Code
Remote environments enable messaging agents from any device, e.g., from phone to laptop.
Letta's Next Phase
Letta announces agents with persistent memory, real computer access, and git-backed memory via Letta Code.
Conversations: Shared Agent Memory Across Concurrent Experiences
Conversations API allows shared memory across parallel user-agent experiences.
Programmatic Tool Calling with Any LLM
Letta API supports programmatic tool calling for any LLM, enabling agent-generated workflows.
Letta Code: A Memory-First Coding Agent
Letta Code is a memory-first coding agent, top on Terminal-Bench benchmark.
Rearchitecting Letta’s Agent Loop: Lessons from ReAct, MemGPT, & Claude Code
New agent architecture optimized for frontier reasoning models.
Letta Evals: Evaluating Agents That Learn
Open-source evaluation framework for testing stateful agents.
Introducing Claude Sonnet 4.5 and the Memory Omni-Tool in Letta
Letta agents can use Sonnet 4.5's memory tool for dynamic memory block management.
Viability Score
How likely is Letta to still be operational in 12 months? Based on 4 signals — momentum (how recently it shipped), wrapper dependency, revenue model, and web presence.
Last calculated: July 2026
How we score →Key Features
- Persistent memory with git-based versioning
- Sleep-time compute for offline learning
- Context Constitution for governing agent context
- Continual learning in token space
- Programmatic context management
- Mods extension system for agent self-improvement
- Conversations API for shared memory across concurrent experiences
- Multi-channel support (Slack, Telegram, Discord, WhatsApp, Signal)
- Schedules and cron jobs for automation
- Remote environments and managed state
- BYOK (bring your own API keys)
- Letta Auto model selection
- npm installable agent package (Letta Code CLI)
- Desktop app for macOS/Windows/Linux
- Web interface (chat.letta.com)
About Letta
Letta is an AI research lab from San Francisco building machines that learn—autonomous agents with persistent memory, continuous self-improvement, and evolving identity. Founded by the creators of MemGPT from UC Berkeley's Sky Computing Lab, Letta produces the Letta Agent, a production-ready system whose memory, skills, and capabilities grow over time through offline learning. Key innovations include Context Constitution for governing agent context management, Context Repositories with git-based memory versioning, sleep-time compute for offline reflection, and continual learning in token space. The agent ships as an npm package (Letta Code CLI), desktop app, and web interface (chat.letta.com). In 2026, Letta introduced Mods—harness-level adaptations that allow agents to self-improve—and the Letta Code app for locally-run, personalized agents. The Conversations API enables shared memory across concurrent experiences. Unlike static models, Letta agents learn during idle time and manage context programmatically, making them ideal for long-running, memory-intensive tasks such as personal assistants, coding agents, and customer support bots that remember every interaction. Compared to tools like AutoGPT or LangChain, Letta prioritizes persistent memory and continual learning over orchestration or prompt chaining.
Behind the Verdict
Letta is one of the most ambitious projects in the agent space—driven by the original MemGPT team, it brings research-grade memory management to production. The git-backed memory and sleep-time compute are genuinely novel; no other tool lets agents reflect offline and evolve their knowledge base over time. The Mods extension system, announced in mid-2026, adds a layer of self-improvement that could differentiate it further. Where does it shine? Any scenario demanding long-running, stateful agents—personal assistants that learn your habits, coding agents that remember past project decisions, customer support bots that track every interaction. The multi-channel support (Slack, Telegram, Discord, WhatsApp, Signal) makes it practical for real-world deployments. Where it bites: this is not a plug-and-play solution. Free accounts are limited to three agents with managed state. Even Pro ($20/mo) caps active agents at 20. Heavy coding users will quickly exceed quota and pay overage—power users often spend $100-200+/mo plus external coding plan costs. Non-technical users will struggle with setup; the CLI/TUI and npm package require comfort with terminals. Compared to AutoGPT or LangChain, Letta is tighter and more opinionated—you get memory infrastructure out of the box, but less flexibility in orchestration. For teams willing to invest engineering time, it's a compelling choice. For one-shot QA or simple chatbots, skip it—there are far simpler tools.
Researching Letta? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Letta actually fits — and what changes day-one when you adopt it.
You install Letta Code via npm, create an agent with your own OpenAI API key, and configure it to learn your project conventions. The agent uses sleep-time compute to analyze your commit history and improve its suggestions over weeks.
Outcome: You get a coding assistant that understands your codebase and coding style, reducing manual context-switching.
You deploy a Letta agent integrated with Slack and your helpdesk. The agent remembers each customer's previous interactions and uses context repositories to access updated product docs, resolving issues faster.
Outcome: First-contact resolution rate improves as the agent learns from each interaction and avoids repeating mistakes.
Use Cases
- Build a coaching agent that remembers user goals, progress, and preferences across months.
- Ship a customer-service agent that recognizes repeat customers and previous conversations.
- Research how main-plus-external context compares to pure RAG on long-conversation benchmarks.
- Prototype a companion app with an agent that accumulates real relationship history.
- Create a coding agent that learns your project conventions over time via Letta Code.
- Run a persistent research assistant that builds knowledge from ongoing queries.
- Automate business workflows with scheduled agents that learn from each run.
- Develop a personal AI assistant that adapts to your behavior and preferences.
Models Under the Hood
as of 2026-07-14
Limitations
- Letta Agent runs on your own API keys from providers like OpenAI, Anthropic, or OpenRouter; inference cost is borne by you.
- The open-source version runs without a login, but cloud features (chat.letta.com, remote environments, managed state, Letta Auto inference) require signing in and may have limits (free accounts support up to 3 agents with managed state).
- Desktop app is available for macOS, Windows, Linux; CLI and web interfaces exist.
as of 2026-06-24
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Letta tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
Developers experimenting with Letta locally using their own API keys, limited to 3 agents on Constellation.
What this tier adds
Free tier is the starting point; you bring your own API keys and get limited Letta Auto usage, up to 3 stateful agents.
Pro
$20/mo
Ideal for
Individual developers running personal agents with moderate usage, up to 20 stateful agents.
What this tier adds
Adds weekly + monthly Letta Auto quota, pay-as-you-go overage, and support for up to 20 agents compared to Free's 3.
API Plan
$20/mo
Ideal for
Teams building automated workflows on the Letta API with unlimited agents and usage-based billing.
What this tier adds
Unlimited agents, API key authentication, pay-as-you-go LLM usage, and $0.10/active agent/mo fee.
Enterprise
Custom
Ideal for
Large organizations needing volume-based pricing, RBAC, SSO, and dedicated support.
What this tier adds
Custom pricing with increased quotas, role-based access control, SAML/OIDC SSO, and dedicated support.
Where the pricing makes sense
The company stage and team size where Letta's pricing actually pencils out — and where peers do it cheaper.
Letta's pricing is developer-friendly with a generous free tier for local use. The Pro tier at $20/mo is competitive for personal AI agents, but heavy users (especially coders) will face additional costs for Letta Auto or external coding plans. Compared to tools like LangChain (free) or OpenAI Assistants (per-token), Letta's managed state and memory features justify the cost for serious agent builders.
Setup time & first value
How long it actually takes to get something useful out of Letta — broken out by persona, not the marketing-page minute.
For developers using local models or BYOK, setup takes 5-10 minutes via npm install and CLI. Creating an account and using Constellation features adds about 5 minutes. Advanced configurations (multi-channel, schedules) may take 1-2 hours.
Switching to or from Letta
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From OpenAI Assistants: You can migrate by reimplementing your existing assistant logic as a Letta agent with programmatic context management.
- ↗To LangChain: You can export agent memory logs and reimplement workflows using LangChain's memory modules.
Integrations
Resources & Guides
- Resourceletta.com
Letta
Making machines that learn. Create stateful agents that remember everything, learn continuously, and improve themselves over time.
- Resourceletta.com
Pricing
Try Letta Code for free
- Resourceletta.com
Blog
Helpful link from letta.com
- Resourcedocs.letta.com
Letta Code
The memory-first coding agent that remembers and learns
Tutorials & Learning
Official links
Popular in Automation & Agents
Presto Voice
Drive-thru voice AI automation for QSR chains to boost revenue and efficiency.
Locus Robotics
Flexible AMR warehouse automation with Physical AI for autonomous fulfillment.
Frequently Asked Questions
Categories
Best-of guides
Used Letta? Help shape our editorial sentiment research.