Stateful Agents Meetup: Building CLI / Coding Agents with Memory
- Date
- 2025-11-21
- Location
- San Francisco, CA, USA
- Host
- Letta Meetup
About this event
Stateful coding agents are moving from demos to real tools, and the hard part is no longer getting an LLM to respond, it is getting an agent to remember, recover context, and stay useful across a real workflow. This meetup is for people who want to build CLI and coding agents that feel dependable in practice, not just clever in a single prompt. About the Event This is an in-person meetup in San Francisco focused on building stateful agents for developer workflows, especially CLI agents and coding agents with memory. The theme is practical: how to design agents that can carry context forward, manage tasks over time, and work inside the kinds of environments developers actually use. Expect a room full of people thinking seriously about agent behavior, memory architectures, tooling, and developer experience. Whether you are experimenting with lightweight command-line assistants or more ambitious code-oriented systems, the discussion is centered on what makes these agents reliable, usable, and worth coming back to. The format is built around learning from peers and having grounded conversations. Rather than treating agents as abstract theory, the meetup is about implementation choices, tradeoffs, and patterns that matter when you are shipping or prototyping something real. What to Expect You can expect a focused evening with technical conversation, shared examples, and opportunities to meet others working on similar problems. The emphasis is on how state actually works in agent systems: memory, task continuity, tool use, context handling, and developer-facing interfaces. Topics that are likely to come up include: Designing memory systems for coding and CLI agents Managing long-running context without overwhelming the model Choosing what an agent should remember, summarize, or discard Building agent workflows around files, repos, terminals, and codebases Evaluating usefulness, reliability, and iteration speed Developer experience questions, including when a CLI is the right interface Because this is a meetup, there is room for both structured content and informal discussion. You should come prepared to hear how other builders are approaching similar challenges, ask detailed questions, and compare notes on what is working and what keeps breaking. If you are early in the process, this is a good place to sharpen your mental model. If you are already building, it is a chance to pressure-test your assumptions with people who understand the technical edge cases. Why Attend If you have worked with agent systems, you already know that memory is where many products become either genuinely useful or frustratingly brittle. Stateless prompting can take you surprisingly far, but coding and CLI workflows often require continuity: remembering prior steps, preserving intent, tracking changes, and operating across sessions. This meetup is valuable because it focuses on that gap between a model that can answer a question and an agent that can support real work. You will be around people interested in the same core problem: how to make agents more coherent, persistent, and effective in developer environments. You should leave with clearer ideas about: What “stateful” means in practice for coding agents Common design tradeoffs around memory and context management Patterns for agent interaction in terminals and code workflows Better questions to ask when designing agent behavior New connections with builders, engineers, and researchers in the space Even if your current project is not strictly a CLI tool, the lessons transfer. Terminal-based workflows force clarity around interfaces, feedback loops, and control, which makes them a strong lens for thinking about agent design more broadly. Practical Details This meetup takes place in person in San Francisco, USA on Thursday, November 20 at 5:30 PM PST. It is best suited for attendees who want live conversation, technical exchange, and the energy of being in a room with others actively building around LLMs and agents. Because the subject is specific, you will get more out of the event if you come with a point of view, a prototype, or at least a concrete curiosity. You do not need to arrive with polished work, but it helps to have a use case, design question, or implementation challenge in mind. A good way to prepare is to think about your own experience with agent memory: Where does your current system lose context? What should the agent remember across steps or sessions? How much autonomy is actually useful in a CLI or coding environment? What would make the tool trustworthy enough to use repeatedly? If those questions sound familiar, this meetup is likely to be worth your time. It is a strong fit for people who want substance over hype and who care about building AI agents that can hold context, assist with code, and operate more like real tools than one-off chat interfaces.
Who should attend
This is for people who want sharper technical conversations about agent design, memory, and developer workflows. - You are building or prototyping **AI agents**, especially tools that operate in the **terminal, editor, or codebase**. - You care about **state, memory, and context management** and want to move beyond stateless prompt-response patterns. - You are an engineer, researcher, or product-minded builder exploring how **LLMs can support real coding workflows**. - You have hit practical issues like context loss, brittle behavior, poor tool use, or unclear agent handoffs and want to compare approaches with others. - You are curious about **CLI interfaces for agents** and want to understand when they are powerful, where they break, and how to make them usable. - You enjoy meetups where the conversation is concrete and technical, and you would rather discuss design tradeoffs and implementation details than broad AI trends. If you are the kind of person sketching memory systems, testing coding agents, or asking how an assistant can stay useful across a session instead of a single prompt, you will likely feel at home here.