TrueFoundry Reduces AI Agent Costs by Up to 75%

TrueFoundry has announced the release of its open-source agent harness, TrueForge, which reduces the operational costs of enterprise AI agents by approximately 30% based on the same model, and by up to about 75% when applying open models. TrueForge is described as a runtime that combines model invocation, tool integration, sandbox execution, approval processes, context management, and session state to enable large language models (LLMs) to perform actual tasks. TrueForge is available in the form of a chat UI, HTTP API, TypeScript SDK, and embedded UI SDK. According to benchmark results, Claude-managed agents solve an average of 10.7 tasks with a single execution cost of $11.8, while TrueForge solves the same number of tasks under the same conditions at a cost of $8.6 per execution. TrueForge, utilizing GLM-5.2, solved an average of 11.7 tasks with a single execution cost of $3.0. TrueFoundry explained that the average was calculated based on n=3 repeated measurements under the same tool and system prompt conditions. The cost difference is influenced not only by model pricing but also by the efficiency of the agent loop. TrueForge recorded over 1,900 stars and 129 forks on GitHub shortly after its release, indicating initial interest from developers. However, it remains uncertain whether this will lead to revenue conversion or widespread corporate adoption.