Conference Presentation
Sara Du, Founder, Asari: MCP & the Future of AI Agents
- The speaker projects that adoption of Model Context Protocol (MCP) will accelerate for the remainder of the year, with a prediction that key players including OpenAI, Microsoft, and a third entity have already committed to supporting the standard.
- While the protocol is currently characterized as being in very early stages with limited app support, the speaker attributes recent progress to Anthropic solving prior design challenges, enabling the potential for critical mass.
- Anticipated applications prioritize aiding users with legacy, "clunky" software environments such as NetSuite and Salesforce, with the expectation that multi-agent systems will soon require formal management frameworks for agents functioning as workforce components.
- New business models are predicted to emerge where entities function primarily as MCP servers to generate significant scale without traditional web application development, drawing a parallel to the impact APIs had on companies like Stripe, Scale, and Twilio.
- Strategic planning for enterprises is expected to begin immediately regarding security infrastructure, specifically involving security scans, whitelisting, and the creation of internal MCP registries to mitigate risks associated with open-source servers, such as credential theft via hidden prompts.
- The outlook identifies significant growth in security protocols for MCPs and the discovery of emergent use cases as the technology matures, while cautioning that users should only deploy use cases where non-deterministic results are acceptable.
- The speaker notes that once introduced to the acronym, the concept of MCP becomes difficult to unsee, and the current presentation is structured to last 17 minutes to explain the protocol and the future of agents.