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.
Agentic AI Foundation Adds Four New Open-Source Tools - OpenSmartRoute
Agentic AI Foundation Adds Four New Open-Source Tools
The Agentic AI Foundation now maintains ten open-source projects. It includes standardized APIs, skill formats, and sandbox environments donated by major companies.
Key points
The total number of AAIF projects is now ten.
Open Responses standardizes API templates for model access.
Agent Skills standardizes how AI agents use external resources.
OpenEnv and OpenSandbox provide tools for training sandboxes.
Why it matters: Standardizing APIs and tools reduces code friction and lowers the risk of hallucinations.
By OpenSmartRoute editorial · written through the router by writer-small
From SiliconANGLE AI - “Agentic AI Foundation adds four more open-source AI tools”
The Agentic AI Foundation Announces Four New Projects
The Agentic AI Foundation added four new open-source tools today. These tools come from major companies like OpenAI and Anthropic. The group now manages ten total projects. This is a big increase from the three it started with.
The foundation is a consortium that keeps open-source AI tools safe. Developers use these tools to build applications without starting from scratch. The new projects solve common problems in building AI agents. They make it easier to move code between different systems.
Open Responses Standardizes AI Model APIs
Open Responses is the first of the new projects. It provides a standardized template for AI model APIs. Before this, every provider used a different API format. This meant software could not move easily between providers.
OpenAI donated this technology to the foundation. They modeled it after their own API system. Developers can now move code between Open Responses and other APIs. They do not need to make major code changes. This speeds up switching between different AI providers.
Many developers already know OpenAI's API well. This familiarity helps them adopt the new standard quickly. The standard works in both directions. Code written for Open Responses works on OpenAI's API too.
Anthropic Donates Agent Skills for Standardized Resources
Anthropic donated a project called Agent Skills. This project provides a standardized format for AI agent skills. Skills are collections of resources that improve an agent's output. These resources can be instructions or automation scripts.
Packaging skills in this standard format makes sharing easier. Teams do not need to change file structures. They can connect skills to workloads immediately. Consistent data organization lowers the risk of errors.
Errors in AI often come from hallucinations. Hallucinations are when an AI makes things up. Standardized skills help agents process data correctly. This reduces the chance of the AI generating false information.
David Soria Parra is an Anthropic engineer. He said the foundation gave the Model Context Protocol a neutral home. He also noted the group continues to invest in open-source tools. This support helps builders everywhere create better AI systems.
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.
Inception AI launched Mercury 3, which is up to eight times faster than GPT-6 Luna and Claude Haiku 5.5. The model uses a new diffusion approach to generate text at 1,068 tokens per second.
Hugging Face and Alibaba Contribute Sandbox Tools
Hugging Face and Alibaba donated two sandbox tools. Sandboxes are technology environments that simulate production infrastructure. They let teams test models safely before going live.
OpenEnv reduces the work of switching training systems. Training systems manage the infrastructure for AI models. OpenEnv automates repetitive management tasks for these systems. This saves time for developers managing complex setups.
OpenSandbox provides APIs for managing sandboxes. These APIs help connect agents inside sandboxes to external apps. This makes testing in a simulated environment much smoother. Teams can verify their agents work correctly before deployment.
AI labs use sandboxes to train their models. For example, a company might simulate developer workstations. This ensures the model behaves correctly in real-world scenarios. The new tools coordinate these sandboxes effectively.
Why Standardization Matters for Model and Agent Workflows
Standardization removes the friction of switching AI providers. Without standards, moving code between systems is hard work. Developers waste time rewriting code for every new provider.
These new projects create a common language for AI tools. They replace the need for custom adapters between systems. This allows teams to focus on building features instead of plumbing.
Consistency in formats helps teams scale their operations. Different teams can share skills and models without friction. This accelerates innovation across the entire organization.
Safety also improves with these standards. Standardized formats reduce the risk of data leaks or errors. Teams can trust that their agents will behave predictably. This is crucial for handling sensitive user data.
The foundation brings together diverse companies to solve these issues. OpenAI, Anthropic, Hugging Face, and Alibaba are all contributors. Their combined expertise creates robust tools for the community.
How OpenSmartRoute Helps Teams Adopt These Standards
Teams using OpenSmartRoute gain immediate benefits from these new standards. The router sends each request to the best-fit model or skill. It uses the standardized formats defined by the new projects.
The router scores every candidate on quality, cost, speed, and safety. The team sets the weights for each request. This ensures the best choice for every specific task.
Hard rules pin requests to specific constraints. For example, text with personal data stays on an on-premises model. The router never crosses a region or cost cap. This guarantees compliance and security for sensitive data.
A new model is just one catalogue entry in the system. It competes on the next request without changing the app. The router learns from outcomes automatically. Models that answer well get more traffic. Failing models get less traffic.
The router works with any OpenAI-compatible provider. It supports open-weight models served locally. It also handles MCP tools and A2A agents. This flexibility lets teams use the new standards immediately.
The savings ledger shows what each routed request cost. It compares this to the cost of the most expensive model. Teams see exactly how much they save by routing correctly.
An input guard spots prompt injection and personal data. It prevents these issues before a request leaves the system. This adds a layer of security that is essential for production use.
The 'osr eval' tool measures routing accuracy on team prompts. It can fail a build if the accuracy drops. This ensures the routing logic remains reliable over time.
The hosted platform keeps a models catalogue with prices. It uses public rankings built from real traffic data. Teams can see which models perform best in their specific context.
What to Do Next with New Open-Source Tools
Teams should start by evaluating their current API usage. They need to understand how much code is locked to one provider. This helps them see the potential savings from standardization.
Check if your team has skills that need a standardized format. Agent Skills can make sharing these resources much easier. Start by packaging your most used skills in the new format.
Consider setting up a sandbox environment for testing. OpenEnv and OpenSandbox can help manage these environments efficiently. Test your agents in the sandbox before moving them to production.
Integrate the new standards into your existing workflow. The Agentic AI Foundation provides clear documentation for each tool. Follow the guides to implement the standards correctly in your stack.
Monitor your costs and performance after making changes. Use the savings ledger to track how much you save. Adjust your weights and rules based on the data you collect.
Keep an eye on the Agentic AI Foundation's updates. New projects may be added to the catalogue soon. Stay informed to ensure you are using the latest tools.
The foundation has done important work in giving tools a neutral home. This neutral home allows everyone to benefit from open-source innovation. Join the community to help shape the future of AI development.
The Agentic AI Foundation Announces Four New Projects
The Agentic AI Foundation added four new projects today. OpenAI, Anthropic, Hugging Face, and Alibaba donated these tools. The total number of maintained projects is now ten. This is up from three when the group launched last December.
The foundation groups open-source tools for AI agents. These tools help developers build and run agent systems. Previously, teams had to manage many different standards separately. Now, they have a single place to find compatible tools.
Open Responses Standardizes AI Model APIs
Open Responses provides a standard template for AI model APIs. OpenAI donated this project to the foundation. The template allows code to move between different APIs easily. Developers do not need to rewrite their code for every provider.
OpenAI modeled this standard after its own API. This means OpenAI workloads can move to other APIs based on the standard. Many developers already know OpenAI APIs well. This familiarity should help teams adopt the new standard quickly.
Anthropic Donates Agent Skills for Standardized Resources
Anthropic donated Agent Skills to the Agentic AI Foundation. This project provides a standard format for AI agent skills. A skill is a collection of resources that improve an agent's output. These resources include instructions, examples, and automation scripts.
Packaging skills in a standard format makes them easy to share. Teams do not need to change file structures before connecting them. Consistent data organization lowers the risk of AI hallucinations.
David Soria Parra, an Anthropic engineer, commented on the donation. He noted the foundation gave the Model Context Protocol a neutral home. Anthropic is committed to continuing to invest in open-source agentic tooling.
Hugging Face and Alibaba Contribute Sandbox Tools
Hugging Face and Alibaba donated two sandbox tools today. OpenEnv comes from Hugging Face. OpenSandbox comes from Alibaba.
Sandboxes simulate the infrastructure for AI models in production. They help teams test agents before running them live. OpenEnv reduces the work involved in switching training systems. OpenSandbox provides APIs for managing these sandboxes.
These tools help coordinate sandboxes used in AI projects. Training systems automate repetitive management tasks. Connecting agents inside sandboxes to external applications becomes easier.
Why Standardization Matters for Model and Agent Workflows
Standardization reduces the friction of switching between AI providers. Without standards, moving code between APIs requires extensive changes. Teams waste time rewriting code for every new provider.
Standardized formats make it easier to share skills and tools. Developers can reuse code across different projects and teams. This increases the speed of development and reduces errors.
Neutral homes for tools prevent vendor lock-in. Teams are not forced to stick with one company forever. They can switch providers without losing their investments. This gives teams more control over their technology stack.
How OpenSmartRoute Helps Teams Adopt These Standards
Teams using OpenSmartRoute gain immediate benefits from these new standards. The router supports Open Responses for standardized APIs. It also handles Agent Skills for consistent skill packaging.
The router works with any OpenAI-compatible provider. This flexibility allows teams to use the new standards immediately. They do not need to change their existing setup.
OpenSmartRoute integrates with the Model Context Protocol. This protocol is now maintained by the Agentic AI Foundation. Teams can use MCP tools without writing custom adapters.
The platform supports A2A agents as well. This allows for standardized communication between different agent systems. Teams can build complex workflows using these open standards.
The savings ledger tracks costs for each routed request. It compares this to the cost of the most expensive model. Teams see exactly how much they save by routing correctly.
An input guard spots prompt injection and personal data. It prevents these issues before a request leaves the system. This adds a layer of security for production use.
The 'osr eval' tool measures routing accuracy on team prompts. It can fail a build if the accuracy drops. This ensures the routing logic remains reliable over time.
The hosted platform keeps a models catalogue with prices. It uses public rankings built from real traffic data. Teams can see which models perform best in their specific context.
What to Do Next with New Open-Source Tools
Teams should start by evaluating their current API usage. They need to understand how much code is locked to one provider. This helps them see the potential savings from standardization.
Check if your team has skills that need a standardized format. Agent Skills can make sharing these resources much easier. Start by packaging your most used skills in the new format.
Consider setting up a sandbox environment for testing. OpenEnv and OpenSandbox can help manage these environments efficiently. Test your agents in the sandbox before moving them to production.
Integrate the new standards into your existing workflow. The Agentic AI Foundation provides clear documentation for each tool. Follow the guides to implement the standards correctly in your stack.
Monitor your costs and performance after making changes. Use the savings ledger to track how much you save. Adjust your weights and rules based on the data you collect.
Keep an eye on the Agentic AI Foundation's updates. New projects may be added to the catalogue soon. Stay informed to ensure you are using the latest tools.
The foundation has done important work in giving tools a neutral home. This neutral home allows everyone to benefit from open-source innovation. Join the community to help shape the future of AI development.
How OpenSmartRoute helps
Teams using OpenSmartRoute gain immediate access to the new standardized APIs and skill formats from the Agentic AI Foundation. The router already scores candidates on quality, cost, speed, and safety, so these new standards improve those scores automatically. Hard rules now pin requests to on-premises models for personal data, ensuring safety without extra setup.
The platform learns from outcomes, so models that answer well get more traffic while failing ones get less. A new model is just one catalogue entry that competes on the next request. Nothing else changes in the app, so teams can adopt these tools instantly.
The hosted platform keeps a models catalogue with prices and public rankings built from real traffic. A savings ledger shows what each routed request cost next to the most expensive model would have cost. Input guards spot prompt injection and personal data before a request leaves the system.
IRONSCALES finds that 77% of organizations report AI tools acted on malicious content. The report links this to faster adoption than security controls.
