TrueFoundry releases TrueForge agent harness to cut AI costs

TrueFoundry recently launched TrueForge, an open-source agent harness that helps developers create and operate AI agents across different model providers. This new tool provides a flexible alternative to managed services, allowing teams to use various large language models while keeping full control over their underlying infrastructure.

Reducing operational expenses for AI development

The primary appeal of TrueForge lies in its potential to significantly decrease the financial burden of running complex AI workflows. TrueFoundry claims that its new harness can lower total operating costs by 50 percent or more. These savings come from the ability to swap expensive proprietary models for more affordable alternatives based on the specific requirements of a task.

Performance and price benchmarks

In a recent test involving 14 enterprise-level tasks, TrueFoundry compared TrueForge against Claude Managed Agents. Both systems performed similarly when using the Anthropic Opus 4.8 model, successfully completing 11 tasks. However, the TrueForge setup cost $8.50 per run, while the managed service cost $11.80, representing a 30 percent price difference for the same level of performance.

The savings became even more pronounced when developers switched to different model configurations. By using the GLM-5.2 model within the TrueForge harness, the cost per run dropped to $2.90. This configuration achieved the same task completion rate as the more expensive managed service, resulting in a 75 percent reduction in total expenditure. These figures provide a compelling case for organizations looking to scale their AI operations without linear cost increases.

Independent validation and scaling

It is important to note that TrueFoundry conducted these benchmarks internally. While the initial data looks promising, the industry still awaits independent validation across larger, more diverse production workloads. Real-world results often vary based on the specific complexity of the prompts and the frequency of tool calls.

Infrastructure costs also play a role in the total cost of ownership. While the software itself is open source, companies must account for the servers and engineering time required to maintain the environment. For many, the trade-off between lower token costs and higher maintenance effort is a central part of the decision-making process.

Strategic control and vendor independence

Beyond immediate cost savings, TrueForge addresses a growing concern among IT managers regarding vendor lock-in. By separating the agent runtime from the model provider, enterprises gain the freedom to move between different AI technologies without rewriting their entire application stack. This modular approach ensures that companies are not beholden to a single provider’s pricing or availability.

Flexibility in model selection

TrueForge supports a wide variety of models, including those from OpenAI and Anthropic, along with over 20 other options. This diversity allows developers to utilize the Model Context Protocol (MCP) to integrate their own servers and API keys. A team can use a high-end model for complex reasoning and a smaller, faster model for routine data processing within the same agent framework.

This flexibility is a major advantage for companies that want to optimize performance. Instead of being restricted to the ecosystem of one provider, they can pick the best tool for every specific sub-task. The open-source nature of the project means that as new models hit the market, they can be integrated into the existing harness quickly.

Governance and regulatory compliance

For organizations in highly regulated sectors like finance or healthcare, control over data and observability is non-negotiable. TrueForge allows these companies to implement their own security protocols and budget caps. Rather than trusting a third-party vendor to manage access and logging, the enterprise keeps everything within its own perimeter.

This self-hosted model provides a level of visibility that managed services often lack. IT departments can monitor every interaction between the agent and the external tools it uses. They can also ensure that the entire environment complies with specific regional data protection laws, such as GDPR or HIPAA, by keeping data processing local.

The future of agent infrastructure

As AI agents become more prevalent in corporate environments, the software layers that manage them are becoming a distinct category of infrastructure. TrueForge represents an early step toward a future where agent harnesses are as standardized as container orchestration tools. However, the industry is still in its infancy, and several hurdles remain before full standardization is achieved.

Challenges to standardization

Industry experts note that we have not yet reached a point of universal agreement on how runtimes should handle context or tool usage. For an infrastructure layer to become truly model-agnostic, there needs to be a consensus on safety requirements and communication protocols. While standards like MCP are helping, the landscape remains fragmented with various vendors competing to define the rules of the road.

Competing platforms often try to differentiate themselves by adding unique features that go beyond basic runtime management. This competition is healthy for innovation but can make it harder for developers to find a single, unified way of working. Despite these challenges, the move toward open-source solutions like TrueForge suggests a desire for more transparent and interoperable systems.

High-volume versus low-volume workloads

The decision to adopt a self-hosted harness often depends on the scale of the operation. For high-volume agents that run continuously, the ability to route tasks to cheaper models can lead to massive savings. In these scenarios, the overhead of managing the infrastructure is easily offset by the reduction in token fees.

Conversely, for smaller projects or unpredictable workloads, managed services might still be the better choice. These services handle all the operational heavy lifting, allowing small teams to focus entirely on building their application. As the technology matures, the line between these two approaches will likely blur, but for now, TrueForge offers a powerful option for those ready to manage their own AI destiny.