Letta is live · free plan, paid from $20/mo · last verified 22 Oct 2026 · viability healthy
Letta
Letta builds stateful AI agents with persistent memory that survives sessions, model swaps, and machines.
If memory is the product you're building, Letta is the most serious stack available, and git-style Context Repositories plus sleep-time compute are bets most orchestration frameworks still haven't made. Budget honestly: casual coding runs ~$100/mo+ in usage per Letta's own docs, and the free tier caps you at 3 stateful agents. For connector-heavy no-code builds, a mainstream agent builder will get you further faster.
Verdict generatedVerified 7d ago · liveness 82/100 · cite: rightaichoice.com/tools/letta
- Teams building long-running agents where remembering past sessions is the core product
- Coding agents that need persistent context across sessions, machines, and model swaps
- Support bots that must recall every prior interaction
- Researchers working on continual learning, memory models, and sleep-time compute
- One-shot chatbots or simple RAG Q&A where persistent memory adds no value
- Teams wanting a plug-and-play no-code agent builder with a broad connector catalog
- Solo developers who need to share agents with colleagues — that requires Teams Pro at $20/seat/mo
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 want a no-code agent builder with prebuilt SaaS connectors and a flat monthly bill, rather than a memory-first platform whose usage costs scale with how much your agent actually runs.
Letta Auto quota resets weekly and monthly, and going past it bills at pay-as-you-go API rates — Letta's docs put casual coding at roughly $100/mo+ and power users at $200+/mo.
At $20/month Pro and $20/seat/month Teams Pro, Letta is seat-cheap but usage-expensive: the real bill is Letta Auto overage at API rates, which Letta prices at ~$100/mo+ for casual coding and $200+/mo for power users. The $20/month API Plan swaps quotas for $0.10 per active agent per month plus $0.00015 per second of tool execution, which suits automated fleets better than humans.
In short
Letta — Letta builds stateful AI agents with persistent memory that survives sessions, model swaps, and machines. Best for Teams building long-running agents where remembering past sessions is the core product, Coding agents that need persistent context across sessions, machines, and model swaps, Support bots that must recall every prior interaction. Free to start; paid plans from $20/mo.
No head-to-head comparison for Letta yet — the alternatives below are the closest matches we list.
What is Letta?
Letta — Letta builds stateful AI agents with persistent memory that survives sessions, model swaps, and machines. Best for Teams building long-running agents where remembering past sessions is the core product, Coding agents that need persistent context across sessions, machines, and model swaps, Support bots that must recall every prior interaction. Free to start; paid plans from $20/mo.
Letta is an AI research lab in San Francisco building what it calls machines that learn: stateful agents whose memory and identity carry across sessions rather than resetting on every chat. The lineage runs back to MemGPT, the 2023 virtual-context paper from UC Berkeley's Sky Computing Lab, and ships today as Letta Agent, a CLI installable via npm i -g @letta-ai/letta-code, plus a desktop app and a web app at chat.letta.com. The audience is developers and teams whose product depends on the agent remembering what happened last time. The research roadmap is the pitch, and it lands in the product in waves. Context Repositories (Feb 2026) rebuild agent memory on git-style versioning so knowledge can be tracked, branched, and reverted. Sleep-time compute has agents reason about context during idle time instead of at inference. The Context Constitution (Apr 2026) sets principles for how an agent manages its own context, and Memory Models (Jun 2026) train memory-native RL for learning that sticks. FrontierMemory (Oct 2026) benchmarks continual learning by measuring how agents improve with memory. On the engineering side, the Agents SDK released August 2026 targets stateful agents that keep identity across models, machines, and interfaces, and September 2026 added dynamic workflows so Letta Code can orchestrate subagents across multiple models and cloud computers. Practical plumbing is there too: a Conversations API sharing memory across concurrent experiences, schedules and cron jobs, remote sandboxes, channels on Slack, Telegram, Discord, WhatsApp, and Signal, and BYOK on every plan via /connect. Pricing splits personal from developer use. Free is $0/month with BYOK and a cap of 3 stateful agents. Pro is $20/month with Letta Auto quota, up to 20 agents, and remote sandboxes; Teams Pro is $20/seat/month for shared agents and access control. The API Plan at $20/month bills $0.10 per active agent per month plus $0.00015 per second of tool execution. Against
Key features
- Persistent agent memory with git-based versioning via Context Repositories
- Memory models trained with memory-native RL for continual learning
- Sleep-time compute: agents reason about context during idle time
- Context Constitution defining how agents manage their own context
- Agents SDK (Aug 2026) for stateful agents that keep identity across models and machines
- Dynamic workflows (Sep 2026) orchestrating subagents across multiple models and cloud computers
- Letta Code CLI installable via npm i -g @letta-ai/letta-code
- Desktop app and web app at chat.letta.com
- Conversations API sharing memory across concurrent experiences
- Channels: Slack, Telegram, Discord, WhatsApp and Signal
- Schedules and cron jobs for automated agent runs
- Remote sandboxes and cloud computers for agent execution
- Self-hosting via Bring Your Own Machine
- Bring your own API keys (BYOK) on every plan via /connect in the CLI
- Letta Auto model selector with Auto, Auto Chat and Auto Fast routes
Behind the verdict
Pick Letta when an agent forgetting yesterday is a bug you can't ship. Long-running coding agents, support bots that must recall every prior interaction, and research work on continual learning are the natural fits; the Agents SDK (Aug 2026) and dynamic workflows (Sep 2026) mean you're not stuck in a demo. We'd reach for it any time the pitch is 'our agent gets better with use' and you want to actually instrument that. Pass when persistence buys you nothing. One-shot chatbots and simple RAG Q&A don't need git-versioned memory, and you'll pay attention and dollars for machinery you never touch. Teams wanting a drag-and-drop builder with a deep connector catalog should look at a mainstream agent builder instead, because Letta's channels cover Slack, Telegram, Discord, WhatsApp, and Signal and not much beyond. The cost is the part buyers misjudge. Chat users often stay inside Pro's quota, but Letta's own docs put casual coding at ~$100/mo+ and power users at or above $200+/mo, and they recommend pairing the CLI with an external coding plan like ChatGPT Plus/Pro or the zAI plan for heavy work. Treat the $20 Pro tier as a starting point, not a ceiling, and set auto-top-up deliberately. Team sharing is gated. Adding members to your organization requires Teams Pro at $20/seat/month, so a solo developer who wants a colleague to poke at an agent has to buy a seat. The API Plan is a different animal entirely, unlimited agents billed at $0.10 per active agent per month plus $0.00015 per second of tool execution, which suits automated workloads and confuses anyone comparing it to Pro line by line. Where it bites: Letta is a research lab first, and the surface area shows it. CLI, desktop, web, SDK, API plans, channels, and self-hosting are a lot of doors, and the docs are dense
Models under the hood
as of 2026-10-23
Limitations
- Letta is memory-first research software, and the constraints follow from that.
- The free tier caps you at 3 stateful agents, and adding teammates to an organization or setting access control requires a Teams Pro account.
- Personal Plans use OAuth through the CLI or web app; Developer Plans use API keys and pure usage-based credit pricing, so the two are not interchangeable in how you authenticate.
- Usage, not seats, drives cost: Letta's docs put casual coding at roughly $100/mo+ and power users at or above $200+/mo, with overage billed at pay-as-you-go API rates once your weekly or monthly Letta Auto quota is gone.
- For heavy coding, Letta recommends combining the CLI with an external coding plan such as ChatGPT Plus/Pro or zAI.
- Documentation — the SDK handbook, CLI reference and quickstart — lives off the marketing site, so expect to read rather than follow a guided setup.
as of 2026-10-18
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.
Is Letta still active?
Letta is live · free plan, paid from $20/mo · last verified 22 Oct 2026 · viability healthy.
Viability score
How well maintained and how widely used is Letta? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What's new in Letta
Checked 4 days agoAcross the latest 3 updates: 1 feature update, 1 launch and 1 news mention.
Letta Code adds dynamic workflows for cloud agents
Letta Code now supports dynamic workflows, letting agents programmatically orchestrate subagents across multiple models and computers, including cloud agents.
Letta Agents SDK released for stateful agents
A new SDK targets stateful, persistent agents that retain identity, memory and experience across models, machines and interfaces.
Memory Models: Towards Agents That Learn
Letta published research on powering agent learning through memory models trained with memory-native RL.
What do users say about Letta?
What people actually say about Letta — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
ResearchedDiscussed by users across 5 sources (Hacker News, YouTube, Stack Overflow, GitHub, Lemmy) · researched Oct 14, 2026.
Weighted by the 71 posts each of 5 sources contributed.
- +Community consistently treats Letta as the reference implementation for persistent agent memory
- +Letta Code hit #1 OSS on Terminal-Bench and drew strong developer enthusiasm
- +Git-style Context Repositories let you track, branch and revert agent memory
- +Sleep-time compute (backed by an arXiv preprint) lets agents learn during idle time
- +Maintainers closed every reported GitHub issue in the dataset — demonstrably responsive
- −Token cost of resending memory every turn is flagged as expensive by users
- −Memory architecture moves complexity to retrieval rather than eliminating it
- −Docs-heavy setup; Stack Overflow users call it more complicated than needed
- −Cloud system-message and tool_rules persistence bugs had to be reported by users
- −Self-hosting and data locality on desktop are not obvious or one-click
- • Token costs from resending memory every turn are the real expense users flag — BYOK means you own that bill
- • Usage-based API plan can scale unpredictably for long-running or high-frequency agents
- • Sleep-time compute is designed to save inference, but any cloud sandbox or remote agent usage adds onto the plan
- • Additional seats in Teams Pro stack up quickly for small teams who only need occasional access
How we verified this
Every listed fact about Letta— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 22 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified Letta 22 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed pricing page
- — changed summary, description, pricing tiers, who should skip it, our verdict, features, who it suits
- — changed switching in, limitations, description, who should skip it, switching out, pricing tiers, summary, our verdict, who it suits, features, setup time, pricing fit, hidden costs, use cases
- — changed who should skip it, features, our verdict, summary, description, pricing tiers
- — changed limitations, pricing page
- — changed SSL certificate
Showing the 6 most recent of 22 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does Letta cost in 2026?
Letta: Free to start; paid plans from $20/mo. Pricing last verified 22 Oct 2026.
Plans compared
Every published Letta tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Free | Free | — | A solo developer testing whether persistent agent memory fits their idea, using their own model keys.Adds: Free entry point: $0/month, BYOK via /connect, but capped at 3 stateful agents. |
| Pro | $20/mo | monthly | An individual running a handful of personal or side-project agents who wants hosted model access without managing keys.Adds: Adds a Letta Auto weekly and monthly quota, pay-as-you-go overage, remote sandboxes, and raises the cap to 20 stateful agents. |
| API Plan | $20/mo | monthly | A small team building automated, high-agent-count workloads on the Letta API with API keys.Adds: Swaps agent caps and quotas for usage-based credit pricing: unlimited agents at $0.10 per active agent per month and $0.00015 per second of tool execution. |
| Teams Pro | $20/seat/mo | monthly, per seat | A team that needs to share agents across colleagues and control who can access them.Adds: At $20/seat/month, adds teammates to your organization, shared agents, access control and permissions, plus org agents that communicate with each other. |
| Enterprise | Custom | quote | A high-volume organization that needs SSO, role-based access and support commitments.Adds: Custom volume pricing with increased quotas, role-based access control, SAML/OIDC SSO and dedicated support, with optional managed per-seat pricing. |
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.
Where the pricing makes sense
The company stage and team size where Letta's pricing actually pencils out — and where peers do it cheaper.
At $20/month Pro and $20/seat/month Teams Pro, Letta is seat-cheap but usage-expensive: the real bill is Letta Auto overage at API rates, which Letta prices at ~$100/mo+ for casual coding and $200+/mo for power users. The $20/month API Plan swaps quotas for $0.10 per active agent per month plus $0.00015 per second of tool execution, which suits automated fleets better than humans.
Has Letta's pricing changed?
No plan price for Letta has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick Letta?
Pick Letta if you are Teams building long-running agents where remembering past sessions is the core product, Coding agents that need persistent context across sessions, machines, and model swaps, Support bots that must recall every prior interaction, Researchers working on continual learning, memory models, and sleep-time compute.
Look elsewhere if you are One-shot chatbots or simple RAG Q&A where persistent memory adds no value, Teams wanting a plug-and-play no-code agent builder with a broad connector catalog, Solo developers who need to share agents with colleagues — that requires Teams Pro at $20/seat/mo, Buyers on a strict flat monthly budget: Letta's docs put casual coding at ~$100/mo+ in usage.
Skip it if Skip Letta if you want a no-code agent builder with prebuilt SaaS connectors and a flat monthly bill, rather than a memory-first platform whose usage costs scale with how much your agent actually runs.
GeneratedReal-world workflow fit
Concrete scenarios for the personas Letta actually fits — and what changes day-one when you adopt it.
You wire the Conversations API and Slack channel so a returning customer's prior tickets live in the agent's memory rather than a pasted transcript, then branch the memory repo when you change the escalation policy.
Outcome: Customers get answers that reference their actual history, and you can revert a memory change that made replies worse.
You install with npm i -g @letta-ai/letta-code, connect your own ChatGPT coding plan via /connect, and let the agent accumulate your repo's conventions across sessions and model swaps.
Outcome: BYOK keeps usage on your provider account, and the agent's project knowledge survives the next model you switch to.
You stand up agents on the $20/month API Plan with unlimited agents and scheduled cron runs, then compare memory-model behaviour over repeated task cycles.
Outcome: You get reproducible stateful agents at $0.10 per active agent per month instead of a per-seat contract.
Use cases
- Build a coaching agent that remembers user goals, progress and preferences across months of sessions.
- Ship a support agent that recognises returning customers and recalls prior conversations without a pasted transcript.
- Run a coding agent via Letta Code that accumulates your project's conventions over time.
- Orchestrate subagents across multiple models and cloud computers with dynamic workflows.
- Prototype a companion app whose agent builds genuine relationship history with each user.
- Schedule agents on cron-style runs so recurring business workflows improve from each execution.
- Run a research assistant that keeps a durable knowledge base across ongoing queries.
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.
CLI: minutes to install from npm and connect keys via /connect. Web app: immediate — the free tier allows 3 stateful agents. Teams: add collaborators and configure permissions, but note membership requires Teams Pro. Production agents on the API Plan need SDK reading first, since the handbook and quickstart live off the marketing site.
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 a pasted chat transcript: move prior context into an agent's memory block so continuity stops depending on the prompt.
- →From a stateless orchestration framework: wrap existing tool calls as Letta tools and let the Agents SDK hold identity across model swaps.
- →From ChatGPT Plus/Pro or zAI coding plans: connect them via /connect for BYOK rather than replacing them.
- →From a custom vector-store memory: replace retrieval-only recall with Context Repositories versioning.
- ↗To a no-code agent builder: export tool definitions and rebuild connectors visually, accepting loss of git-style memory versioning.
- ↗To a raw model API: reimplement memory and scheduling yourself; Letta's value is the memory layer you would be dropping.
- ↗To a stateless framework: keep the prompts, lose the persistent identity, and paste context back in per run.
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
YouTube returned 6 videos for “Letta”, and we withheld 6: 6 could not be judged, because “Letta” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Letta.
Official links
What are the best Letta alternatives?
The closest alternatives to Letta we list are Cognee, Mem0 and Athena Public. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Cognee | Freemium | Pricing not published | — |
| Mem0 | Freemium | Pricing not published | — |
| Athena Public | Free | Free | — |
Alternatives to Letta
View all
Cognee
Cognee gives AI agents persistent knowledge-graph memory with cited recall across sessions, teams, and tools.
Mem0
Mem0 is a memory layer that gives AI agents and apps persistent context across sessions and users
Athena Public
Athena Public gives your LLM a persistent, editable memory layer that survives across sessions and IDEs.
Tools that pair well with Letta
Common stack mates teams adopt alongside Letta, with the specific reason each pairing earns its keep.
Cognee
Cognee gives AI agents persistent knowledge-graph memory with cited recall across sessions, teams, and tools.
Mem0
Mem0 is a memory layer that gives AI agents and apps persistent context across sessions and users
Athena Public
Athena Public gives your LLM a persistent, editable memory layer that survives across sessions and IDEs.
Frequently asked questions
Categories
Used Letta? Help shape our editorial sentiment research.