Teradata adds context engine and execution layer to Tera agents

Teradata is adding a context engine, an execution layer called Tera Harness, and reusable agent skills to its Tera AI workspace. In internal SWE-bench Pro tests, Tera used 73% fewer tokens than Claude Code, finished tasks 42% faster, and cut costs 58% on the same Opus 5 model. General availability is planned for December.

• Tera Harness plans execution before LLM calls and applies 84 execution patterns
• On SWE-bench Pro: 73% fewer tokens and 58% lower cost than Claude Code
• Tasks completed 42% faster with higher completion rates
• General availability of the new capabilities planned for December