Letta Office Hours: Mods, Signal Support, GLM-5.2, and Slack Agents

Weekly Letta Office Hours with Cameron from the Letta team. This week: mods, Signal support, dedicated GLM access, local agents vs. Constellation agents, Letta Desktop onboarding, context window sizing, and why Letta’s Slack channels are a more flexible, multiplayer version of “Claude Tag.” Cameron starts with product updates: the new mods registry and one-week mod challenge, Signal as a Letta Code channel, a dedicated `letta/glm` endpoint for GLM-5.2, support for multiple Codex plans, improved worktree behavior, Kimi K2.7 in the model registry, and local secret manager support. The second half is community Q&A covering skills vs. mods, how mods persist in the harness, local vs. Constellation agents, memory portability, context window tradeoffs, self-hosting changes, the Letta Agent SDK, app server architecture, and Slack-based agent collaboration. Office hours happen Thursdays at 11:30 AM Pacific. Recordings are posted here for anyone who can’t make it live. Join the Discord:   / discord   Chapters: 00:00 Intro and office hours format 01:05 Mods registry and what mods can do 04:10 One-week mod challenge 06:59 Signal support for Letta Code channels 10:07 Dedicated GLM-5.2 endpoint 12:09 Multiple Codex plans 13:24 Worktree improvements 13:55 Kimi K2.7 in the model registry 14:29 Local secret manager support 15:27 Q&A begins 16:30 Community mods and secret scanning 17:07 Local mode, privacy, and telemetry 17:46 Image support for GLM-5.2 18:44 Kimi thinking leaks and model feedback 21:57 Are mods the same in Desktop and CLI? 22:31 Community projects and 3D printing agents 24:58 Kimi K2.7, Base10, and model hosting tradeoffs 26:20 Discord channels, roles, and Letta Code naming 29:31 Context Doctor improvements 32:21 Skills vs. mods 36:30 Personalization, “user” language, and harness prompting 44:52 Letta Code commit footers and AI disclosure 53:55 Updating older agent system prompts 54:31 Desktop onboarding, local agents, and Constellation 59:42 Moving local agents to Constellation 01:01:36 Benefits of Constellation agents 01:03:09 Local vs. Constellation UX confusion 01:05:21 Why run local agents? 01:07:42 Compaction and long-running conversations 01:09:19 Do mods live in MemFS? 01:10:30 Are mods one-time or persistent? 01:11:43 Context window sizing and 128k defaults 01:17:33 Mods across local and Railway environments 01:19:20 Self-hosting, Docker deprecation, and local mode 01:21:03 Multi-user access and channels 01:21:51 App server and Letta Agent SDK 01:23:33 Claude Tag vs. Letta Slack channels 01:26:36 Memory transparency and Claude Tag limitations 01:28:16 Permissions, compliance, and agents-as-coworkers 01:30:00 Multiplayer agent teams and model agnosticism Topics: Letta, Letta Code, Letta Desktop, AI agents, agent memory, MemFS, mods, skills, Signal, Slack channels, Constellation, local agents, GLM-5.2, Kimi K2.7, context windows, agent SDK, Claude Tag.