> Markdown version of [/videos/100272-rest-in-peace-why-llms-can-t-crud](https://www.wearedevelopers.com/videos/100272-rest-in-peace-why-llms-can-t-crud). 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).

---

# REST In Peace: Why LLMs Can't CRUD

Why do AI agents fail at simple CRUD operations? Traditional REST APIs force probabilistic LLMs into rigid schemas. Discover how intent-based APIs finally bridge this machine communication gap.

- **Speakers:** [Martin Sakowski](https://www.wearedevelopers.com/@martin-sakowski), [Martin Karrer](https://www.wearedevelopers.com/@martin-karrer)
- **Event:** World Congress 2026 Europe
- **Published:** July 10, 2026
- **Duration:** 30:40
- **URL:** https://www.wearedevelopers.com/videos/100272-rest-in-peace-why-llms-can-t-crud

## Summary

As enterprise software moves from supportive chatbots to action-driven AI agents, current system architectures are hitting a "co-pilot plateau." Traditional REST APIs were designed for human developers and microservices, creating severe friction when LLMs attempt to execute downstream operations. Agents frequently mismanage standard CRUD operations because they struggle with strict schema rigidity, inefficient token-heavy payloads, and complex orchestration demands. Furthermore, even a 99% success rate on a single API execution rapidly plummets due to error compounding when an agent must piece together dozens of granular, stateful requests just to complete an everyday task.

To bridge this communication gap, systems must evolve toward intent-based APIs hosted on MCP (Model Context Protocol) servers. Rather than forcing probabilistic LLMs to parse irrelevant data and reverse-engineer rollback logic for human-centric 409 Conflict errors, developers must move structural "plumbing" back into backend microservices. Within this new paradigm, the agent focuses purely on "judgment"—passing high-level goals, constraints, and preferences—while the system handles the workflow transactions natively. This shift mirrors the design of highly effective developer CLIs, mapping highly complex internal operations into unified, intent-rich access points.

Ultimately, layering intent-based APIs over existing REST infrastructure dramatically scales performance, improves token economics, and provides machine-actionable error handling for autonomous agents. This specific architecture also strengthens overall system security by introducing "valet key" authentication, granting scoped, time-boxed access to an agent rather than relying on broad API "house keys." Transitioning to this model empowers developer and people-ops teams to redefine how they track success, focusing purely on absolute agent task success rates rather than granular stateless metrics.

**Keywords:** intent-based APIs, LLM agent orchestration, REST API friction, Model Context Protocol, stateful AI workflows, error compounding, token economics, machine-actionable error handling, valet key authentication, API schema rigidity, agent task success rate, co-pilot plateau, backend orchestration abstraction, autonomous agent deployment, generative AI API design

## Chapters

1. **Evaluating AI agent performance on daily assistance tasks** (00:03) — Benchmarking pure AI assistants against standard API integrations reveals a major gap in task execution reliability.
1. **Why standard REST APIs fail autonomous AI agents** (02:33) — Traditional HTTP-based patterns rely on human developer intuition and do not scale to computational agentic workflows.
1. **Orchestration explosion and error compounding in workflows** (04:51) — Generating and managing multi-step API sequences creates cascading failures due to probabilistic model execution limits.
1. **API parsing inefficiencies and schema rigidity limits** (07:03) — Exposing large static resource objects forces models to consume excessive tokens and hallucinate exact schema matches.
1. **Navigating unhelpful errors and implicit API documentation** (09:08) — Status codes block automated error recovery while implicit design rules prevent agents from correctly deducing field semantics.
1. **Resolving agent challenges with intent-based API architectures** (13:10) — Abstracting technical implementations through outcome-focused interfaces eliminates orchestration overhead and pushes business logic handling to the system.
1. **Building an intent layer over legacy REST APIs** (18:48) — Adopting a backend-for-frontend pattern allows standardized intent handling without requiring a complete rewrite of traditional systems.
1. **Securing agent access through scoped valet keys** (20:14) — Authenticating agent behaviors through time-boxed and scoped operational limits prevents unauthorized execution across the larger resource surface.
1. **Tracking task success metrics and optimizing agent environments** (22:04) — Measuring operation completion rather than individual status codes yields accurate diagnostics for token cost and latency reduction.
1. **The paradigm shift toward intent-based system communication** (24:22) — Adapting model context protocols to wrap backend intent paves the way for scalable and resilient autonomous environments.
1. **Evaluating alternatives like HATEOAS and GraphQL** (27:43) — Analyzing HATEOAS and GraphQL reveals that theoretically sound API specifications often fail agents due to adoption gaps or query complexity.

## Related Moments

- [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")
- [Adapting legacy system architectures for agent autonomy](https://www.wearedevelopers.com/videos/100278-designing-for-agents-will-make-you-better-at-designing-for-humans) (from "Designing for Agents Will Make You Better at Designing for Humans")
- [Redefining the software architect role for AI pipelines](https://www.wearedevelopers.com/videos/100190-architecture-3-0-from-90-to-99-999-reliability-in-building-ai-systems) (from "Architecture 3.0: From 90% to 99.999% Reliability in Building AI Systems")
- [Integrating intent-based code generation and agent implementation](https://www.wearedevelopers.com/videos/1855-the-intent-engineer-closing-the-gap-between-business-engineering-manuel-klein) (from "The Intent Engineer: Closing the Gap Between Business & Engineering - Manuel Klein")
- [Building agentic workflows using prompt engineering and language models](https://www.wearedevelopers.com/videos/1266-navigating-the-ai-revolution-in-software-development) (from "Navigating the AI Revolution in Software Development")

## Related Articles

- [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)
- [Never delegate the understanding](https://www.wearedevelopers.com/magazine/749-never-delegate-the-understanding)
- [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering)

## Related Jobs

- [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**
- [Senior Backend Developer — AI: MCP & Agent Engine](https://www.wearedevelopers.com/jobs/48297-senior-backend-developer-ai-mcp-agent-engine) at **basebox GmbH**
- [Staff Software Engineer, Copilot Experiences](https://www.wearedevelopers.com/jobs/ext/164361-staff-software-engineer-copilot-experiences) at **GitHub**
- [Principal Product Manager, Agent Platform](https://www.wearedevelopers.com/jobs/ext/277541-principal-product-manager-agent-platform) at **GitHub**
- [Principal Field Architect - AI Agents](https://www.wearedevelopers.com/jobs/ext/1442858-principal-field-architect-ai-agents) at **Twilio**
- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**