Selector Foundry brings Git workflows and incident replay to network AI agents

Selector launched Foundry, a development and runtime environment for building, testing, versioning and governing AI agents in network operations. Agents run on Pydantic AI, are committed to the customer's Git repo, reviewed via pull requests and replayed against historical incidents before production, with one-step rollback.

• Agents run on Pydantic AI rather than heavier frameworks like CrewAI
• Before production, agents are replayed against the customer's past incidents
• Guardrails cap token usage and constrain what an agent may conclude
• Full automation is the goal, but the transition will be gradual