At the recent MCP Summit in New York City, The New Stack sat down with AWS’s Luca Chang, who works
The post How AWS Bedrock is shaping Model Context Protocol appeared first on The New Stack.
At the recent MCP Summit in New York City, The New Stack sat down with AWS’s Luca Chang, who works on the Bedrock team and is an MCP Specification Maintainer to boot.
MCP has quickly become the industry standard for connecting AI models and agents to tools and data, making it a critical component of the broader artificial intelligence stack. Chang explained how developers are choosing what to improve in MCP, how Amazon decided what to contribute to the open-source tool, and who is paying whose token budgets!
When TNS sat down with Chang, the developer was fresh out of a pre-conference MCP maintainer meeting. How does the MCP developer pool decide what to build next for the protocol? By bringing together a diverse set of priorities and perspectives, and hashing out what needs to be added. Don’t worry that the model will yield groupthink. According to Chang, the MCP maintainer cohort is sufficiently broad that there were too many topics to fit into the group’s recent meeting.
Not that making choices is easy. Chang explained that maintainers’ work includes weighing problems that need to be solved against smaller MCP changes that could unlock capabilities that are “really creative and interesting.” (Your edge case will face real competition for bandwidth.)
This is where corporations can play a role in building out open-source projects. Amazon’s major contributions to MCP (Tasks, Elicitations) arose from the AWS team mapping its cloud products to the protocol and finding that it came up short. So, it helped improve MCP’s ability to interact with cloud computing platforms.
At AWS, Chang explained, “we don’t exactly look to make contributions [to MCP] as quickly as possible […] the contributions sort of fall out of our customer use cases […] once we explore a use case and see that there is [a] gap in the protocol itself. That’s when we say, ‘oh, there might be something here that we can give back to the community.’”
For more on where Chang sees the most market demand for MCP servers, and why he doesn’t want agent-specific MCP servers, check out this latest episode of The New Stack podcast.
Group Created with Sketch.
| # | Наименование новости | Тональность | Информативность | Дата публикации |
|---|---|---|---|---|
| 1 | As agentic AI explodes, Amazon doubles down on MCP | 0 | 11.43 | 16-04-2026 |
| 2 | Why the Linux Foundation adopted MCP, with Jim Zemlin and Mazin Gilbert | 0 | 12.89 | 06-05-2026 |
| 3 | Cut AI token usage by 96%? Here’s how AWS Strands Agents does it. | 0 | 14.09 | 29-04-2026 |
| 4 | Five AI rivals just backed a shared plugin standard. Here’s why it matters for developers. | 0 | 16.61 | 08-08-2026 |
| 5 | How context rot drags down AI and LLM results for enterprises, and how to fix it | 0 | 12.14 | 09-03-2026 |
| 6 | Why your company should (try to) build its own AI SRE | 0 | 15.19 | 30-07-2026 |
| 7 | The blank-check AI coding era is dead. Here’s what comes next. | 0 | 16.51 | 06-08-2026 |
| 8 | Your AI agent’s next tool call may be valid but wrong. AWS’s Dogwood promises to fix that. | 0 | 16.08 | 06-08-2026 |
| 9 | With the launch of Meko, Yugabyte targets the data layer that’s breaking multi-agent AI systems | 0 | 18.8 | 07-05-2026 |
| 10 | Your AI strategy is built on layers of API sediment | 0 | 14.93 | 17-02-2026 |