Urgent.News

What's breaking now, across thousands of outlets.

AI

TrueFoundry's open source AI agent harness TrueForge boasts 30%-75% cheaper task completion than Claude Managed Agents

Another day, another new AI agent harness is released. Only this time, it's one that aims to solve a growing enterprise problem as AI agents proliferate: enabling greater developer control of agents and tools, while reducing cost. TrueFoundry , a San Francisco B2B machine learning startup co-founded in 2021 by former Meta and Google engineers, has released its own custom TrueForge harness under…

TrueFoundry's open source AI agent harness TrueForge boasts 30%-75% cheaper task completion than Claude Managed Agents

TrueFoundry, a San Francisco-based machine learning startup, has released its open-source AI agent harness called TrueForge under the permissive MIT License on GitHub. This harness is designed to enable greater developer control of AI agents and tools while reducing costs. TrueForge can be used with any preferred AI models, forked, modified, self-hosted, and incorporated into commercial products.

In a benchmark test using GLM-5.2 LLM, TrueForge paired with the open-source model successfully completed 11 out of 14 tasks on DevRev's Enterprise-Bench. The cost for these tasks was 75% less than completing the same tasks with Anthropic's Claude Managed Agents, costing $2.90 compared to $11.80. When using the same model, TrueForge still achieved a cost savings of roughly 30% ($8.50 vs $11.80) compared to Claude Managed Agents.

TrueFoundry's co-founder and COO, Anuraag Gutgutia, stated that customer requests drove the development of this harness. The primary difference between TrueForge and other harnesses, like cloud-managed or commercial-provider-managed ones, is that TrueForge enables vendor-neutral use of managed agents and allows enterprises to use them in a cost-effective manner.

TrueForge's architecture focuses on context engineering, which aims to reduce waste in AI agent operations. This includes delaying the loading of MCP tool schemas, delegating isolated tasks to subagents, moving oversized tool results to files, processing structured results through code, and automatically compacting long-running conversations. These features help reduce unnecessary compute and allow more agents to run concurrently on a single server.

TrueForge is designed to move from local development to a shared production runtime while keeping models, tools, and infrastructure interchangeable. The company emphasizes that TrueForge is not a replacement for other harnesses but serves as a vendor-neutral way to use AI agents at a lower cost. TrueFoundry also offers a paid AI Gateway that provides central control over model and MCP access, credentials, permissions, budgets, and observability.

Written by urgent.news from VentureBeat's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at venturebeat.com →

More in AI

From Flat Logs to Execution Trees: Debugging Modern AI Agents

An agent trace is usually written as a sequence of events because append-only data is simple to produce: span_started span_started span_ended span_started span_ended span_ended Developers do not want…

  • Agent traces appear as a series of events due to append-only data structure simplicity.
  • Developers construct execution trees to visualize relationships between operations.
  • Reliable execution tree requires stable trace and span identity assignment.

More from Wednesday 19 August →