Shrey

LX is a local experimentation platform that helps AI agents build and test task-specific harnesses—the prompts, skills, guardrails and tools that shape how they work. Given sample inputs, a goal and an output contract, LX runs an automated loop where a worker attempts the task and an improver revises the harness while model weights stay fixed. It records trajectories, tool calls, checks, errors and usage so developers and agents can understand failures and assess changes. Through its Web UI, CLI and MCP connectivity, LX supports reproducible experiments and exports task-specific trajectories that can be curated for model fine-tuning.