Comparison

DSPy vs Letta

A like-for-like, spec-level comparison. Both entries are verified against their docs and repo.

DSPy

Declarative, programmable prompt programming from Stanford.

  • + Compiles and optimizes prompts instead of hand-writing them
  • + Strong for repeatable, measurable prompt engineering
  • − Different mental model from typical agent frameworks
  • − Smaller ecosystem of tool integrations

Letta

Stateful agents with long-term memory (formerly MemGPT).

  • + First-class long-term memory and stateful agents
  • + Lets agents persist and self-edit context
  • − Narrower focus than general orchestration frameworks
  • − Memory model adds complexity
Spec DSPy Letta
Role Agent framework Agent framework
Tags agent-framework, library, persona-solo-builder, persona-platform-engineer agent-framework, library, multi-agent, memory, persona-solo-builder, persona-platform-engineer
License MIT Apache-2.0
Open source Yes Yes
Self-hostable Yes Yes
MCP support No Yes
Pricing free freemium
Price Free Free tier
Usage cost BYOK BYOK
Models byok claude, gpt, gemini, byok
Languages python python
GitHub stars 35.8k 23.6k
Last activity 2026-07-02 2026-07-03