Letta Code · General-purpose agent with persistent state
Product and access
Letta Code connects a local agent runtime with persistent agent state. Letta Cloud is the default backend; a local backend can be selected at startup.
Provider
Letta develops Letta Code and the cloud platform for agents with persistent memory. The repository is maintained by the team behind MemGPT.
- Operator
- Letta
- Organization
- Company
- Website
- www.letta.com
Core features
5 features from provider sources · checked on 10 October 2026. Availability and limitations follow below.
Manage memory in Git
MemFS represents agent memory as an editable Git filesystem with version history.
Source 3Continue agents across interfaces
CLI, desktop and browser access the same agents in the selected Letta account.
Source 1Use persistent skills
Skills can belong to the project, computer or agent memory and provide relevant resources and scripts.
Source 4Schedule tasks
One-time or recurring prompts run locally with an active session or as cloud schedules in a sandbox context.
Source 5Set tool permission levels
standard, acceptEdits, strict and unrestricted control approvals; additional allow and deny rules restrict actions.
Source 6
Requirements and limits
Limitation
Interactive CLI starts with broad autonomy
The interactive CLI defaults to unrestricted. Select standard or strict for individual approvals before shell commands, edits and subagents.
Source 6Limitation
Local schedules need a running process
Local schedules run only while a Letta app, CLI or server process is open. Cloud schedules use UTC and may fall back to the sandbox when the target computer is offline.
Source 5Limitation
Distinguish cloud state and local backups
Cloud agents store memory, identity and conversations with Letta. Users are responsible for backing up local MemFS repositories.
Source 3Limitation
Check quotas and overage
Free includes three stateful agents, Pro up to 20. Letta Auto quotas are described without a specific public token allowance; overage is paid.
Source 2
Operation & oversight
Operational details for this profile have not yet been researched.
Practical tasks
Editorial examples based on documented features. No product tests performed.
Maintain project knowledge across sessions
- Your input
- Project materials, verified facts and rules for persistent memory.
- Possible result
- An agent that reuses saved project knowledge in later tasks.
- What to check
- Check saved facts, memory changes and the backend storage location.
Set up a recurring check
- Your input
- A bounded check, schedule and reachable target computer.
- Possible result
- Recurring check reports in the selected agent context.
- What to check
- Check the running process, time zone, execution target and spending limit.
Project materials, verified facts and rules for persistent memory.
Your possible prompt
Read the project rules. Investigate [error], propose a small change, and run the relevant checks. Show me the diff and any remaining uncertainties.
Getting-started idea based on documented capabilities. Not an executed product test. Adapt the task, data, and permissions.
Setup and pricing
Setup
Install CLI or desktop, select a local or cloud backend and configure model access and tool permissions
Prices and included usage
Pricing checked
Personal plans separate the base price and Letta Auto quota from overage. Your own model keys are billed by the model provider.
- Free
- USD 0/month
Three stateful agents, limited Letta Auto and your own API keys or external coding plans.
Price source - Pro
- USD 20/month
Up to 20 stateful agents and weekly and monthly Letta Auto quota; usage-based overage.
Price source - Teams Pro
- USD 20/seat/month
Shared agents, access control and Letta Auto quota.
Price source - API Plan
- USD 20/month plus usage
Unlimited agents; USD 0.10 per active agent/month, USD 0.00015/second of server-side tool execution and model costs.
Price source
- The specific Letta Auto quota is not publicly quantified; check current allowances in the account.
- The Apache-2.0 runtime and a local backend have their own operating and model costs. Client-side CLI tools incur no Letta CPU charge according to the pricing documentation.
Sources and evidence
Official product documentation reviewed. Workflows and assessments are editorial; hands-on product tests are outside the scope of this directory.
Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.
-
github.com /letta-ai/letta-code
-
docs.letta.com /pricing
-
docs.letta.com /concepts/memfs
-
docs.letta.com /configuration/skills
-
docs.letta.com /configuration/schedules
-
docs.letta.com /configuration/permissions
