# Postman and Baseten: Build Performant Agents, Run Them At Lower Cost

Everyone is building AI agents, but far fewer enterprises run them at scale without costs getting out of hand. We’re excited to partner with [Baseten](https://www.baseten.co/) to fix this problem for enterprises by combining the fastest inference runtimes, cross-cloud high availability, and developer workflows that fit together. Agents powered by the Baseten Inference Stack and the Postman Agent Stack run fast, and with even lower running costs.

## Postman: The Stack for Agents

 Since agents need APIs to know and act, we launched the Agent Stack (Astropods, Fabric, and Passport), in addition to our family of AI-native products, to make any enterprise agentic-ready. Postman is where 40+ million developers build, test, and ship APIs. It gives teams the platform to design agents, connect them to the tools they need, and learn how they perform in the real world. As enterprises build their own internal AI agents using Astro AI, Postman’s production-grade control plane and agent operating system (agentOS), they need a fast, cost-efficient model-serving option. Baseten is now our recommended inference partner for enterprise teams as they move agentic workflows into production. > “Every dollar we save on inference is a dollar we can put back into giving developers more AI capability for free. That’s why we chose to partner with Baseten for our enterprise customers: fast, efficient model serving that cuts inference costs, coupled with deep post-training expertise. That lets us deliver more value back to our community.” Abhinav Asthana, Founder and CEO, Postman

## Baseten: Enterprise AI Infrastructure

 Baseten is the leading inference platform for teams running AI in production. It provides fast, reliable, cost-efficient model serving, with the GPU capacity, post-training infrastructure, and hands-on support to run open-weight and customized models at scale. The Postman team set out to reduce the cost of running Postman Agent Mode, used by over 40 million developers, while holding quality and speed steady. Baseten helps by providing post-training infrastructure to customize open-weight models (thereby lowering costs without sacrificing quality), plus model serving powered by the [Baseten Inference Stack](https://www.baseten.co/resources/guide/the-baseten-inference-stack/). > “Proud to partner with the Postman team to post-train open models that meet the quality bar but are many times cheaper and faster to run.” Amir Haghighat, CTO and Co-founder, Baseten

## Postman + Baseten, together for the enterprise

 Together, Postman and Baseten help enterprises build and run AI agents at scale while keeping costs under control. #### Connect the stacks for building and running agents

 Postman is where teams design agents, connect them to the APIs and tools they need, and see how they perform in production. Baseten is where the models behind those agents run. Through our partnership, teams can go from an agent built in Astro AI to production serving on the Baseten Inference Stack without stitching together infrastructure. You also get the capacity and uptime that only cross-cloud high availability (Baseten’s [MCM system](https://www.baseten.co/blog/how-we-built-multi-cloud-capacity-management/)) can provide, from day one. #### Match the right model to the task

 An agent workflow is made of many steps, like planning, tool selection, API calls, and summarization. Only some of those steps require a frontier model. Teams can route high-volume, simpler steps to smaller open models served on Baseten and save frontier models for the hardest reasoning. Match the model to the task and cut cost without cutting quality. #### Keep learning on your own data

 Real-world usage is the best signal for making agents better. With Baseten's post-training infrastructure, teams can fine-tune open models on real usage, so agents improve over time. #### Serve for agent-scale traffic

 A single agent task can trigger many model calls, and enterprise deployments multiply that across all of their users. By pairing multi-cloud GPU capacity with runtime optimizations in the Baseten Inference Stack, Baseten keeps latency low as request volume climbs or spikes. #### Partner with two applied AI teams

 Both Postman’s and Baseten’s teams work directly with enterprise customers on their hardest problems. Customers get a shared channel with engineers from both partners, with same-day support and hands-on help. ## Get started

 Building agents and watching your token bill? Reach out to build with Postman and Baseten.