# Letta Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/letta
> Markdown URL: https://aitinkerers.org/technologies/letta.md
> Technology record last updated: 2026-02-27T19:11:52Z
> Generated: 2026-09-21T04:44:28Z

Letta is the open-source platform that transforms stateless LLMs into perpetual, self-improving AI agents with persistent, hierarchical memory.

Letta (formerly MemGPT) delivers the AI Operating System for stateful agents: a critical upgrade for LLMs. The platform, originating from UC Berkeley's Sky Computing Lab, utilizes a hierarchical memory model (core in-context blocks and external archival storage) to ensure agents maintain context across multi-session workflows. Developers build, deploy, and manage these stateful agents using robust Python and TypeScript SDKs, or the Agent Development Environment (ADE). This model-agnostic framework supports local providers (Ollama) and cloud services, enabling long-running, self-improving agents for complex use cases like customer support and research assistance.

- Official technology site: https://letta.com
- Public AI Tinkerers demos and talks: 2
- Result page: 1 of 1

## Recent Public Talks and Demos

### [My Coworkers Are Agents: Running a GTM Consultancy Where Slack-Native AI Workers Build, Monitor, and Self-Heal Client Automations](https://seattle.aitinkerers.org/talks/rsvp_VkekCRBY8TQ)

I built a team of AI "employees" — GTM Engineering agents — that run go-to-market operations for real clients at my consultancy, Tarka. Each agent lives in Slack (one channel per client), and end-to-end it builds, operates, monitors, and auto-recovers the client's GTM stack: prospecting in Deepline → CRM in Attio/Clarify → outbound email in Instantly → LinkedIn outreach via allman.sh, all orchestrated through n8n workflows. You message the agent in Slack like a coworker — "pipeline for this ICP is dry" — and it ships the automation, watches it run, and fixes it when a step breaks. Live, I'll show: (1) an agent taking a Slack request and standing up / repairing an n8n workflow across the GTM tools; (2) the architecture of the platform underneath it (ergoly) — how each agent is an isolated worker bound to a single client; and (3) the monitoring + auto-recovery loop that lets these agents run unattended. I'll pull up the real system: the Slack control plane, the worker sandboxes, and the four-tier memory + credential-injection design that keeps every client's data and secrets separate.

- Event context: GTMxAI Engineering July - Seattle Tech Week — 2026-07-30 — Seattle
- Public talk page: https://seattle.aitinkerers.org/talks/rsvp_VkekCRBY8TQ

### [Building stateful agents with Letta](https://sf.aitinkerers.org/talks/rsvp_L0p2bJAj-8I)

In this demo, we introduce the Letta ADE (Agent Development Environment) for building, customizing, and deploying Letta agents, which have long term memory powered by MemGPT.

- Event context: AI Tinkerers - Halloween Edition - October 2024 Meetup — 2024-10-23 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_L0p2bJAj-8I

## Related Technologies

- [Claude](https://aitinkerers.org/technologies/claude) ([Markdown](https://aitinkerers.org/technologies/claude.md)) — 173 public demos
- [Letta ADE](https://aitinkerers.org/technologies/letta-ade) ([Markdown](https://aitinkerers.org/technologies/letta-ade.md)) — 1 public demo
- [MemGPT](https://aitinkerers.org/technologies/memgpt) ([Markdown](https://aitinkerers.org/technologies/memgpt.md)) — 1 public demo
- [mitmproxy](https://aitinkerers.org/technologies/mitmproxy) ([Markdown](https://aitinkerers.org/technologies/mitmproxy.md)) — 2 public demos
- [Next](https://aitinkerers.org/technologies/next) ([Markdown](https://aitinkerers.org/technologies/next.md)) — 186 public demos
- [Supabase Postgres](https://aitinkerers.org/technologies/supabase-postgres) ([Markdown](https://aitinkerers.org/technologies/supabase-postgres.md)) — 2 public demos