AMESA

AMESA, formerly Composabl, is a web platform and API for industrial engineers, control teams, and data scientists to build, train, and deploy multi-agent AI systems that optimize physical and industrial processes.

Key capabilities include:

• No-code Agent Orchestration Studio to design agent workflows by defining goals, constraints, and success criteria, plus visual debugging and benchmarking for iteration.
• Agent Import API and Python SDK/CLI to integrate custom logic such as Python-based controllers, LLMs, ML models, reinforcement learning policies, and other algorithms.
• Simulation-driven training using simulations generated from historical CSV data or Docker-based physics models and emulators, enabling safer validation before production.
• Deployment to real operations by connecting agents to industrial environments and systems (including SCADA, MES, HMI, ERP, robotics, and factory equipment) with decision traceability.

Pricing: Paid, typically delivered via 2 to 3 month pilots that include platform seats, agent deployment, support, and use case implementation.

AMESA stands out for its machine teaching approach that helps domain experts encode operator knowledge into agent behavior, along with published industrial case studies reporting outcomes such as 15% better blending ($400k/day ROI) and 3% improved control on a Siemens gPROMS digital twin ($1.7M/year ROI).