​Giving an agent a task is easy. Letting it run inside a real engineering organization is harder.

​Before agents can work on their own, teams need to answer some practical questions. What can the agent access? How do you keep token usage under control? And how do you give an entire engineering team one shared setup without creating a security mess?

​This meetup brings together three engineers working on those problems now. We will hear about securing agent credentials, reducing token usage and how to build agents for scale.

​For engineers and technical leads building with coding agents. Free, limited seats.

​Agenda
* 17:30–18:00 Welcome
* 18:00–18:20 Talk 1
* 18:20–18:40 Talk 2
* 18:40-19:00 Break
* 19:00–19:20 Talk 3
* 19:20–20:00 Cheers

​Talks

​Talk 1: Stop paying your agents to forget

​Yahav Ohana (Co-Founder Boost @JFrog)

​Agent loops can burn through tokens by repeatedly reading the same files, carrying irrelevant context, and retrying failed approaches.

​Yahav will share practical ways to reduce token usage while keeping agents effective.

​Talk 2: Don't give your agent the keys

​Pablo Goldszeft (Principal Engineer @ Islo, previously at Snyk)

​Agents need access to code, infrastructure, and internal tools to get useful work done. Giving them broad, permanent credentials is the fastest route to a security incident.

​Pablo will cover how to give agents the access they need without handing over the keys.

​Talk 3: The first step toward autonomy is a shared repo

​Tal Aharon (Engineer @ Legit Security)

​At Legit Security, one repository gives 60 engineers access to the same agent toolchain: 27 plugins, 117 skills, and 11 MCP servers.

​Tal will explain how they built it and why guardrails do not limit autonomy. They create the trust needed to give agents more access, more responsibility, and more time to work on their own.