> Markdown version of [/videos/100291-building-apis-for-agents-vs-systems-is-mcp-the-answer?t=1324](https://www.wearedevelopers.com/videos/100291-building-apis-for-agents-vs-systems-is-mcp-the-answer?t=1324). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Building APIs for Agents vs Systems. Is MCP the answer?

Standard REST APIs will make your AI agents hallucinate. Learn how the Model Context Protocol (MCP) shifts orchestration to the server for a flawless agent experience.

- **Speakers:** [Adam Bird](https://www.wearedevelopers.com/@adam-bird)
- **Event:** World Congress 2026 Europe
- **Published:** July 10, 2026
- **Duration:** 27:52
- **URL:** https://www.wearedevelopers.com/videos/100291-building-apis-for-agents-vs-systems-is-mcp-the-answer

## Summary

Decades of REST API design have centered around deterministic systems and predictable orchestration by developers. However, when the API consumer becomes an AI agent—a non-deterministic actor that seeks a final outcome rather than composable primitives—traditional API contracts fracture. Exposing too much granularity overwhelms agents, causing them to hallucinate steps or fail outright when attempting complex workflow orchestrations. 

To address this gap, the Model Context Protocol (MCP) shifts the orchestration burden back from the client to the API server. When optimizing for agent experience (AX) instead of developer experience (DX), providers must drastically abstract complexity, ideally restricting workflows to a maximum of two steps. Under MCP, documentation is indistinguishable from code: parameter names act as literal prompt instructions, and runtime-generated schemas allow APIs to present highly tailored namespaces. For instance, an API connected to an HRIS can dynamically expose "candidate" and "interviewer" terminology exactly matched to the client's localized context, rather than generic resource terms.

Adapting to agentic consumers forces a radical rethinking of established API fundamentals. Standard HTTP error codes must be discarded in favor of natural language instructions that guide agents to autonomously unblock themselves. Furthermore, because older models frequently hallucinate or fabricate execution records, robust API observability becomes a critical asset for support teams. Ultimately, giving agents excessive data permissions is a security liability; providers must fiercely filter data and restrict tools to build trust, acknowledging that agents operate with the unpredictable variations of human users.

**Keywords:** model context protocol (MCP), ai agent API design, non-deterministic API consumers, agent experience (AX), REST API limitations, runtime schema generation, natural language error handling, contextual tool discovery, API observability, data filtering constraints, scheduling API abstractions, LLM tool hallucination, HRIS integration schemas, workflow orchestration, dynamic interface namespaces

## Chapters

1. **Adapting web APIs for deterministic systems versus agents** (00:39) — Designing APIs for agents requires moving past composable REST primitives to deliver isolated outcomes directly.
1. **Managing complex workflows for unpredictable AI agent consumers** (03:30) — While REST relies on composable primitives, complex multi-step interactions cause drift and overwhelm non-deterministic models.
1. **Centralizing execution orchestration on the server using MCP** (08:19) — Managing state and workflows server-side prevents agents from failing complex operational sequences.
1. **Optimizing tool granularity and limiting multi-step automated processes** (09:52) — Reducing operations to a maximum of two procedural steps prevents exponential complexity loops and unexpected execution failures.
1. **Merging documentation and source code using MCP schemas** (12:35) — Defining parameters and intent inline guarantees that documentation perfectly mirrors the actual code structure.
1. **Designing error messages as actionable guides for agents** (14:29) — Replacing static status codes with descriptive, actionable error responses allows agents to unblock themselves and dynamically correct inputs.
1. **Delivering dynamic API footprints and schemas at runtime** (16:24) — Generating specific API capabilities based on runtime context eliminates backwards compatibility bloat and increases execution success rates.
1. **Implementing strict data filtering to maintain security boundaries** (20:00) — Restricting agent access to the minimum necessary application data prevents unpredictable data breaches and preserves customer trust.
1. **Building observability to verify unexpected AI agent behavior** (22:04) — Robust monitoring acts as an essential diagnostic tool when AI models hallucinate or falsely report API execution paths.
1. **Combining deterministic workflows with agent-based protocol automation models** (23:32) — Combining deterministic infrastructure with interaction protocols provides practical guardrails against unpredictable autonomous workflows.
1. **Constraining nested agent communication using domain-specific translation layers** (26:07) — Standardizing domain responsibilities via an intermediate protocol provides defined guardrails that prevent autonomous agents from triggering unintended chain reactions.

## Related Moments

- [Core concepts and advantages of the Model Context Protocol](https://www.wearedevelopers.com/videos/100305-api-mcp-or-mcp-app-choosing-the-right-surface-for-ai-agents) (from "API, MCP or MCP App? Choosing the right surface for AI agents")
- [Treating AI tools as standard governed enterprise capabilities](https://www.wearedevelopers.com/videos/100336-from-apis-to-mcp-enterprise-governance-registry-and-controls) (from "From APIs to MCP: Enterprise Governance, Registry, and Controls")
- [Standardizing agent interactions with the Web MCP proposal](https://www.wearedevelopers.com/videos/1805-wearedevelopers-live-modern-devops-for-iot-devices-and-more) (from "WeAreDevelopers LIVE - Modern DevOps for IoT Devices and More")
- [Understanding foundational concepts of LLMs, agents, and MCPs](https://www.wearedevelopers.com/videos/1768-boost-productivity-with-ai-figma-playwright-mcp-workflows-aris-markogiannakis) (from "Boost Productivity with AI: Figma & Playwright MCP Workflows - Aris Markogiannakis")
- [Defining true agent behavior versus traditional enterprise chatbots](https://www.wearedevelopers.com/videos/1538-composable-intelligence-how-henkel-and-microsoft-are-shaping-the-agent-ecosystem) (from "Composable Intelligence: How Henkel and Microsoft Are Shaping the Agent Ecosystem")
- [Balancing developer autonomy with the adoption of coding agents](https://www.wearedevelopers.com/videos/100198-the-last-mile-of-ai-from-prototype-to-production) (from "The Last Mile of AI: From Prototype to Production")

## Related Articles

- [Everything a Developer Needs to Know About MCP with Neo4j](https://www.wearedevelopers.com/magazine/604-everything-a-developer-needs-to-know-about-mcp-with-neo4j)
- [WebMCP: Empowering Agents as First-Class Citizens of the Web](https://www.wearedevelopers.com/magazine/696-webmcp-empowering-agents-as-first-class-citizens-of-the-web)
- [The Web We Broke (And Why AI Agents Are Paying the Price) - AgentCon Berlin](https://www.wearedevelopers.com/magazine/735-the-web-we-broke-and-why-ai-agents-are-paying-the-price-agentcon-berlin)
- [What is Agentic Programming and Why Should Developers Care?](https://www.wearedevelopers.com/magazine/625-what-is-agentic-programming-and-why-should-developers-care)

## Related Jobs

- [Senior Backend Developer — AI: MCP & Agent Engine](https://www.wearedevelopers.com/jobs/48297-senior-backend-developer-ai-mcp-agent-engine) at **basebox GmbH**
- [Principal Product Manager, Agent Platform](https://www.wearedevelopers.com/jobs/ext/277541-principal-product-manager-agent-platform) at **GitHub**
- [AI Full Stack Engineer](https://www.wearedevelopers.com/jobs/ext/1354435-ai-full-stack-engineer) at **Almedia**
- [Principal Field Architect - AI Agents](https://www.wearedevelopers.com/jobs/ext/1442858-principal-field-architect-ai-agents) at **Twilio**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**