← Knowledge

Letta Office Hours

Timestamped explanatory guides to Letta Office Hours episodes, organized as historical records of the platform’s changing agent architecture.

Letta Office Hours is a public video series about the changing architecture and practice of persistent AI agents. These guides turn each recording into a compact, timestamped explanation of its major announcements, demonstrations, questions, and design connections.

The pages are historical records, not current product manuals. Letta changes quickly; a feature described in an older episode may have been renamed, redesigned, removed, or absorbed into a different surface. The original recording remains the primary source for what was said at that time.

Episodes

2026

2025

  • 2025-12-18: Letta Code Demo, Agent Skills, Claude Code Proxy & More — Office hours on Letta Code, agent skills, persistent memory, sub-agents, and emerging ideas for shared blocks and file-system-like agent environments.

  • 2025-12-04: December 4th, 2025 — December 4, 2025 office hours on the v1 SDK launch, message search, Learning SDK examples, skills, lettactl, personal agents, and Ezra updates.

  • 2025-11-06: November 6th, 2025 — November 6, 2025 office hours on the v1 SDK migration, shared archives, Letta Code improvements, AI Memory SDK v0.2, scheduling, and Ezra.

  • 2025-10-30: October 30th, 2025 — October 30, 2025 office hours on Letta Code, Co, the Obsidian plugin, personal agents, and the platform shift toward terminal-first and note-centric workflows.

  • 2025-10-16: October 16th, 2025 — October 16, 2025 office hours on voice support, the runs viewer, Telegram improvements, stateful-agent patterns, and practical tips for building with memory.

  • 2025-10-02: October 2nd, 2025 — October 2, 2025 office hours on the v1 alpha, memory tools, mobile agents, Obsidian, Bluesky, and the shift toward a new agent architecture.

  • 2025-09-25: September 25th, 2025 — Office hours on the AI SDK v5 update, cloud reliability work, open-source release planning, memory behavior, and model/provider recommendations.

Recurring themes

  • Persistent identity and memory. The series repeatedly separates an agent from any one chat window, model call, or client application.

  • Execution surfaces. Letta Code, Desktop, Chat, channels, sandboxes, schedules, and the Agent SDK appear as different ways to connect durable agents to work.

  • Harness extension. Skills, tools, MCP, mods, protocols, and provider compatibility are treated as different layers rather than synonyms.

  • Inspectable state. Memory files, commits, runs, approvals, traces, and tests make agent behavior reviewable instead of merely fluent.

  • Teaching and evaluation. Later episodes increasingly ask how agents learn from trajectories, specifications, feedback, and explicit release gates.

Public sources

Sources

  1. Letta on YouTube
  2. Letta documentation
  3. Letta on GitHub

Connections

Related

Linked here

Suggest a correction ↗