AI Control: TrueFoundry Offers Free Developer Tool

San Francisco-based TrueFoundry, a startup founded in 2021 by former Meta and Google engineers, has released its TrueForge AI agent harness under the MIT License. This open-source tool empowers developers with greater control over AI agents and tools while aiming to reduce associated costs.

When tested, TrueForge, paired with the GLM-5.2 LLM, completed 11 out of 14 tasks on DevRev's Enterprise-Bench for significantly less than competitor harnesses. TrueFoundry claims cost savings of up to 75% compared to Anthropic's Claude Managed Agents. Even when using the same model, Opus 4.8, TrueForge reportedly offers a 30% cost reduction.

TrueForge operates above TrueFoundry's existing paid AI Gateway, managing the loop for models to reason, call tools, and complete tasks. Developers can start locally and scale to shared deployments using Docker Compose or Helm. The harness focuses on context engineering to minimize waste, such as delaying tool schema loading and processing large results as files.

The company envisions TrueForge complementing existing solutions, allowing enterprises to use it alongside cloud-managed or commercial harnesses. The ultimate goal for TrueFoundry is for its AI Gateway to become the central layer for managing model access, credentials, and budgets, irrespective of the agents or harnesses chosen by the enterprise.

TrueFoundry, which has raised approximately $21 million in funding, previously focused on MLOps before expanding into managing LLM applications. Beta users like NetApp have utilized the harness for incident response and ticket triage, exposing internal agents as self-service tools for developers. The company also acquired Seldon AI in June 2026 to integrate MLOps capabilities.