GPT-5.6 (Sol / Terra / Luna) is now evaluated on TrustVector — with day-1 independent verification, incl. METR's benchmark-cheating findings.

Read the evaluation
Evaluation record · memgpt

Letta (formerly MemGPT)

v0.x

Letta Inc.

Agentmemorylong-contextopen-sourcerebranded
78
Strong
About This Agent

REBRANDED: MemGPT became Letta (Letta Inc., a UC Berkeley spinout) in September 2024; "MemGPT" now refers only to the research technique. Letta is a memory-enhanced LLM agent platform enabling long-term context via OS-inspired virtual context management. In March 2026 Letta announced a pivot to Letta Code, a client-side, model-agnostic agent harness with persistent memory, deprecating server-side memory tools, templates, identities, MCP integrations, and tool rules.

Last Evaluated: July 9, 2026
Official Website

Trust Vector Analysis

Dimension Breakdown

🚀Performance & Reliability
+
memory management

Memory system testing

Evidence
MemGPT Documentation — Hierarchical memory system with virtual context management
highVerified: 2026-07-09
long context handling

Long context testing

Evidence
Research Paper — Enables unlimited context through memory pagination
highVerified: 2026-07-09
conversation continuity

Continuity testing

Evidence
Memory Features — Maintains conversation state across long interactions
highVerified: 2026-07-09
llm integration

LLM integration testing

Evidence
Model Support — Supports OpenAI, Anthropic, local models via backends
highVerified: 2026-07-09
self editing memory

Self-editing capability testing

Evidence
Memory Editing — Agent can edit its own memory structures
mediumVerified: 2026-07-09
latency

Performance benchmarking

Evidence
Performance — Memory management adds overhead to LLM calls
mediumVerified: 2026-07-09
🛡️Security
+
memory isolation

Isolation testing

Evidence
Architecture — Per-agent memory isolation in storage backend
mediumVerified: 2026-07-09
self hosting

Deployment security assessment

Evidence
Deployment — Full self-hosting with pip install or Docker
highVerified: 2026-07-09
data persistence

Data security assessment

Evidence
Storage — Configurable storage backends (PostgreSQL, SQLite)
highVerified: 2026-07-09
open source

Open source assessment

Evidence
GitHub — Apache 2.0 license, 12k+ stars, research-backed
highVerified: 2026-07-09
memory security

Memory security assessment

Evidence
Security Features — Memory security requires careful configuration
mediumVerified: 2026-07-09
🔒Privacy & Compliance
+
memory privacy

Privacy architecture review

Evidence
Memory Storage — Memory stored in configured database backend
mediumVerified: 2026-07-09
gdpr compliance

Compliance capabilities assessment

Evidence
Self-Hosted — GDPR compliance possible with self-hosted deployment
mediumVerified: 2026-07-09
local deployment

Deployment options assessment

Evidence
Installation — Complete local deployment with local LLMs supported
highVerified: 2026-07-09
llm data sharing

Data flow analysis

Evidence
LLM Integration — Memory context sent to configured LLM provider
mediumVerified: 2026-07-09
memory deletion

Data deletion assessment

Evidence
Memory Management — Memory can be deleted, archival system available
mediumVerified: 2026-07-09
👁️Trust & Transparency
+
documentation quality

Documentation completeness review

Evidence
Documentation — Good documentation with research paper backing
highVerified: 2026-07-09
research backed

Research foundation assessment

Evidence
Research Paper — Published research paper explaining memory system
highVerified: 2026-07-09
memory visibility

Transparency assessment

Evidence
Memory Inspection — Agent memory can be inspected and debugged
mediumVerified: 2026-07-09
open source

Open source assessment

Evidence
GitHub — Apache 2.0, 12k+ stars, active community
highVerified: 2026-07-09
community support

Community engagement analysis

Evidence
Community — Active Discord community and GitHub discussions
mediumVerified: 2026-07-09
⚙️Operational Excellence
+
ease of integration

Integration complexity assessment

Evidence
Python Package — Pip install with CLI and Python API
highVerified: 2026-07-09
scalability

Scalability testing

Evidence
Architecture — Scalability depends on database backend and infrastructure
mediumVerified: 2026-07-09
cost predictability

Pricing model analysis

Evidence
Pricing — Free Apache 2.0, costs for LLM API and storage
highVerified: 2026-07-09
monitoring

Monitoring features assessment

Evidence
Monitoring Features — Basic logging, requires external monitoring tools
mediumVerified: 2026-07-09
production readiness

Production readiness assessment

Evidence
Maturity — Active development, production use requires careful setup
Letta Rebranding Announcement — MemGPT rebranded to Letta (Letta Inc., UC Berkeley spinout, Sept 2024); active commercial development continues under the Letta name
Letta's Next Phase (blog) — March 2026 pivot to Letta Code (open-source, client-side, model-agnostic agent harness with git-backed 'MemFS' memory); server-side memory tools, templates, identities, MCP integrations, and tool rules deprecated by mid-April 2026; letta server package latest 0.16.8 (May 2026), repo active (23.7k stars, pushed July 2026)
mediumVerified: 2026-07-09
cli interface

CLI capabilities assessment

Evidence
CLI — Feature-rich CLI for agent management
highVerified: 2026-07-09
Strengths
  • +Breakthrough memory system enabling unlimited conversation context
  • +Research-backed approach (published paper) with novel virtual context
  • +Open source (Apache 2.0) with active community (23k+ stars)
  • +Self-hosting option for complete memory control
  • +Supports multiple LLM providers and local models
  • +Enables truly stateful, personalized agent interactions
Limitations
  • !Memory management adds latency overhead to interactions
  • !Requires database backend setup for persistence
  • !Complexity in managing and debugging agent memory
  • !Active development, some features still experimental
  • !Limited production tooling and monitoring features
  • !Memory pagination can introduce unpredictability
  • !Rebranded to Letta (Sept 2024, older MemGPT docs/packages outdated); March 2026 pivot to client-side Letta Code harness deprecated server-side memory tools, templates, identities, MCP integrations, and tool rules by mid-April 2026, requiring migration for existing server-based deployments
Metadata
license: Apache 2.0
supported models
0: OpenAI
1: Anthropic
2: Azure OpenAI
3: Local LLMs via Ollama
programming languages
0: Python
deployment type: Self-hosted (pip, Docker) or Letta Cloud; Letta Code client-side harness (2026)
tool support
0: Memory management
1: Tool use
2: Custom functions
pricing model: Free open source (Letta Cloud managed offering available)
github stars: 23700+
github repo: https://github.com/letta-ai/letta
latest version: letta 0.16.8 (May 2026); Letta Code released April 2026
first release: 2023
research paper: https://arxiv.org/abs/2310.08560
storage backends: PostgreSQL, SQLite, Chroma
rebranding: Now called Letta (formerly MemGPT); flagship shifting to Letta Code client-side harness as of March 2026
pricing: Free (open source)

Use Case Ratings

customer support

Excellent for long-term customer relationships with memory

code generation

Good for maintaining context across coding sessions

research assistant

Excellent for long-form research with extensive context

data analysis

Good for maintaining analysis context over time

content creation

Good for maintaining style and context in long projects

education

Excellent for personalized tutoring with student history

healthcare

Good for maintaining patient context and history

financial analysis

Useful for maintaining client context and history

legal compliance

Excellent for maintaining context across long documents

creative writing

Good for maintaining creative context across sessions