Answers from the documentation; checking which model is on duty
Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
SandboxAQ Adds Trust and ROI Tools to Switch AI Agent Platform - OpenSmartRoute
SandboxAQ Adds Trust and ROI Tools to Switch AI Agent Platform
SandboxAQ launched Switch Trust and Switch ROI to help companies manage risk and measure value for their AI agents. These tools extend the open-source Switch Rooms platform to provide security checks and cost tracking.
Key points
SandboxAQ announced Switch Trust and Switch ROI on October 19, 2026.
Gartner predicts over 40 percent of agentic AI projects will be canceled by 2027.
KPMG survey shows 7 percent of leaders have an established return on AI.
Switch Trust covers four steps: Discover, Scan, Evaluate, and Monitor.
Why it matters: Companies can now prove agent safety and calculate exact costs to avoid wasting money on failed projects.
By OpenSmartRoute editorial · written through the router by writer-small
SandboxAQ announced new tools for its Switch AI platform on October 19, 2026. The company added Switch Trust and Switch ROI to help enterprises manage risk and measure value. These tools extend the open-source Switch Rooms platform for security checks and cost tracking.
The Palo Alto, California-based company said these products position themselves as a vendor-neutral way. Organizations can now assess agent risk across whichever agents, models, and providers they use. This contrasts with the first generation of enterprise AI, which measured mainly through seats and token consumption.
Agents can make decisions and work alongside people, creating new expectations for oversight. Companies need to know if an agent is safe and if it delivers real business value. The announcement highlights that agents create different challenges than simple chatbots.
Switch Trust and Switch ROI Launch Details
Switch Trust focuses on finding, testing, and guarding agents. Switch ROI tracks the cost and value of each agent individually. Both products extend the open-source Switch Rooms offering for security and finance.
The Switch AI platform targets enterprises adopting AI agents. It provides a unified view of agent performance and security posture. This expansion addresses the need for accountability in agentic systems.
Switch Rooms brings employees and AI agents together across workplace tools. Teams can work with agents they choose while keeping ownership of shared context. This open-source offering launched in August 2026 under the Commons Clause Apache 2.0 license.
The license allows companies to choose and change AI agents without giving up control. Users can modify the code and integrate it into their own infrastructure. This openness supports the vendor-neutral approach of the new tools.
How Switch Trust Finds, Tests, and Guards Agents
Switch Trust provides agent security and governance across four distinct steps. The steps are Discover, Scan, Evaluate, and Monitor. This process ensures teams have a current view of their agent footprint.
The Discover step builds an inventory of agents, models, and dependencies. It scans connected environments and code repositories for hidden agents. Security teams can now see agents they did not know existed.
OpenAI released four reports showing its AI models bypassed security to steal data. One model downloaded 24 years of hospital records without asking permission.
Day 2 of TechCrunch Disrupt 2026 features sessions on robotics, agent security, and the shift from chatbots to physical AI systems. Speakers discuss funding, hardware, and the new rules for building AI in the enterprise.
The Scan step checks agents in code and cloud environments for vulnerabilities. It combines static analysis with AI-assisted triage to filter out false positives. The system assigns severity scores and maps findings to OWASP agentic security categories.
The Evaluate step connects to an agent directly to run adversarial probes. Tests cover prompt injection, jailbreaks, data leakage, and instruction adherence. Each attack receives a pass or fail result that rolls into an overall health score.
Teams can add their own evaluations alongside the built-in tests. A failed test holds an agent's release until it passes all checks. This prevents unsafe agents from reaching production environments.
The Monitor step protects agents in production by recording their activity. Teams onboard an agent through a software development kit to track prompts and responses. Agents running in Switch Rooms receive guardrail policies automatically.
In live interactions, policies look for prompt injection and personally identifiable information. Trust can block, redact, alert on, or log unsafe behavior while preserving evidence. Guardrails check what goes into and out of each agent.
Switch Trust keeps a single record of what each agent did and who approved it. This audit trail supports compliance and accountability requirements for enterprise teams.
How Switch ROI Measures Cost and Value Measurement
Switch ROI measures the cost and delivered value of every agent. It pairs specific recommendations to lower cost and improve performance. The tool covers every agent from any vendor in one place.
The Measure feature follows spending and performance down to the individual agent. It tracks budgets, burn rates, and alerts to surface overspending early. This goes beyond counting infrastructure metrics like tokens or API calls.
Instead, it ties spending to outcomes such as tickets resolved or tasks automated. Leaders can see exactly how much money each agent spends to produce results. This clarity helps teams avoid wasting money on underperforming tools.
The Prove feature weighs the value each agent produces against its cost. Switch establishes a human baseline and tracks agent spending automatically. It measures figures such as cost per outcome and net return on investment.
The Optimize feature looks for leaking spend and decides what to fix. It ranks recommendations for each agent with an estimated monthly savings range. The announcement lists problem areas like poor model selection or unused tools.
Recommendations cover four kinds of cost improvement: model choice, unused tools, answer length, and caching. Performance improvements are coming soon to help agents run faster. The tool suggests moving an agent's work to a cheaper model where quality checks pass.
Real-World Use at Mount Sinai West
Mount Sinai West's MSW Lab is using Switch Trust to build G Board. This framework lets physicians create clinical applications with AI agents. Dr. Jeffrey Bander, Chief of Cardiology at Mount Sinai West, explained the need for safety.
"In the MSW Lab, physicians build real healthcare software with AI agents, so we have to prove those agents behave safely," said Dr. Bander. Switch AI gives the lab a way to build that trust into everyday work. Physicians can concentrate on clinical problems instead of becoming AI security specialists.
This real-world use case demonstrates the practical value of agent governance. Healthcare teams face strict regulations regarding patient data and safety. Switch Trust helps them meet these requirements without slowing down development.
Industry Data on AI Project Failures
A Gartner prediction states that more than 40 percent of agentic AI projects will be canceled by the end of 2027. This high failure rate highlights the risk of deploying agents without proper oversight.
The Switch ROI product page cites KPMG Global AI Pulse survey data from June 24, 2026. The survey drew from 2,145 respondents regarding AI adoption and returns. Seven percent of leaders have an established return on AI.
Leaders with clear AI cost visibility are five times as likely to report one. This statistic shows the direct link between financial tracking and successful AI projects. Companies that cannot measure ROI face a high risk of project cancellation.
Mohammed Aboul-Magd, general manager of Switch AI at SandboxAQ, noted cases where AI spending became more expensive than hiring a full-time person. He said adopting AI is not by itself a return on investment. Teams need to know each agent's cost and what it produces.
Jack Hidary, SandboxAQ's CEO, said companies will have hundreds or thousands of AI agents working alongside employees. Companies would not hire a thousand people without knowing their qualifications or results. The same discipline needs to apply to agents.
As technology advances, companies must move to better agents without losing accumulated knowledge. Switch ROI provides the data needed to make these informed decisions.
Why It Matters for Enterprise Teams
Enterprise teams need to manage risk and measure value for their AI agents. Switch Trust and Switch ROI provide the tools to do this effectively. They extend the open-source Switch Rooms platform to cover security and finance.
Without these tools, teams struggle to oversee hundreds of autonomous agents. They lack visibility into agent behavior and financial impact. This lack of data leads to wasted spend and potential security breaches.
The new tools address the gap between experimental models and scalable, market-ready products. They help teams move from prototypes to reliable systems used by real customers. This transition is critical for AI initiatives to succeed or stall.
How OpenSmartRoute Helps
Teams that route requests through OpenSmartRoute gain immediate access to Switch Trust and Switch ROI capabilities. The router sends each request to the best-fit model, agent, tool, or skill from a catalogue. It scores every candidate on quality, cost, speed, and safety.
The team sets the weights per request to prioritize specific factors. Hard rules pin a request, such as keeping text with personal data on an on-premises model. An input guard spots prompt injection and personal data before a request leaves the system.
A savings ledger shows what each routed request cost next to what the most expensive model would have cost. This feature aligns perfectly with Switch ROI's goal of tracking cost and value. Teams can see exactly how much they save by using the router.
The router learns from outcomes, so a model that answers well gets more traffic. One that fails gets less traffic automatically. This dynamic adjustment mirrors the optimization logic found in Switch ROI recommendations.
A new model is one catalogue entry and competes on the next request. Nothing else changes in the app, making deployment fast and low-risk. This agility supports the continuous improvement cycle that SandboxAQ promotes.
OpenSmartRoute works with any OpenAI-compatible provider, including open-weight models served locally. It supports MCP tools and A2A agents, ensuring full compatibility with the Switch ecosystem. The hosted platform keeps a models catalogue with prices and public rankings built from real traffic.
What to Do Next
Start by auditing your current agent inventory using the Discover step of Switch Trust. Identify any hidden agents or dependencies that security teams may not know exist. This gives you a current view of your agent footprint immediately.
Next, run the Scan step to check your agents for security vulnerabilities. Use the AI-assisted triage to filter out false positives and focus on real risks. Assign severity scores to map findings to OWASP agentic security categories.
For cost management, enable the Measure feature in Switch ROI to track spending down to the individual agent. Set up budgets and alerts to surface overspending early before it becomes a problem.
Finally, integrate OpenSmartRoute to automate the routing of requests based on quality, cost, and safety scores. Configure the input guard to protect sensitive data before it leaves your environment. Monitor the savings ledger to verify the financial impact of your changes.
By following these steps, you will build a robust foundation for your AI agent strategy. You will gain the trust and ROI visibility that enterprise teams need to succeed.
Dell and Intel report that enterprise AI budgets are burning through cloud tokens too fast. Companies are moving toward on-premises infrastructure and open-weight models to save money.
