Cachly vs MemGPT:
Which AI Memory Is Right for Developers?

What MemGPT became

In September 2024, MemGPT became part of Letta. Today's Letta Code is a stateful agent harness: its agents have memory and identity, and they learn by rewriting their own memory and skills. You talk to them from a CLI, a desktop app, chat.letta.com, or Slack, Telegram and Discord. MemFS keeps all their context in git.

Letta Code is open source under Apache-2.0. Letta Cloud is the default backend; the local runtime keeps all agent state on your machine, with no Letta account. You pick the model, from Anthropic and OpenAI to local models.

What cachly is

cachly gives memory to the coding assistant you already use: Claude Code as a plugin, and Cursor, Windsurf, VS Code with GitHub Copilot, Cline, Continue.dev and Zed over MCP. It keeps lessons from fixes and decisions in your codebase: what worked, what failed, the files and the commands. Its data is stored in Germany.

The difference

With Letta, the agent is the product. Its memory belongs to that agent. Claude Code or Cursor can hand work to a Letta agent through Letta's hosted MCP server and read its reply.

With cachly, your assistant stays the same. It keeps working as before, and cachly adds what the last session learned to the next one. In Claude Code, the UserPromptSubmit hook adds the matching lessons to every prompt.

When Letta is the right choice

• You want an agent with its own memory and identity that learns over long stretches of time.
• You want to reach the same agent from a terminal, a desktop app, a browser or Slack.
• You want its memory as files in git, in your own repository.
• You want to choose the model yourself and run everything on your own machine.

When cachly is the right choice

• You keep your coding assistant and want it to remember fixes, dead ends and the files involved.
• You want matching lessons next to each prompt, without asking for them.
• Your data should be stored in Germany, with embeddings computed in the EU.

Side by side, with sources

The full comparison covers memory, MCP, models, hosting, teams, license and pricing, with a source for every row: cachly vs Letta.

A correction: earlier versions of this post said Letta had no MCP integration and a free tier only when self-hosted. Both were wrong. Letta runs a hosted MCP server, its agents call MCP tools, and Letta Cloud has a free plan with your own API keys.

More comparisons

Side by side, with a source for every row: cachly vs Supermemory · cachly vs Zep · cachly vs claude-mem · cachly vs Basic Memory · cachly vs mem0 · cachly vs Letta.