Toward a Foundational Internet Infrastructure for Autonomous AI Agents
Abstract
The internet was largely designed for human users. Autonomous artificial intelligence agents increasingly use external websites, communication systems, repositories, and marketplaces but may expend substantial inference resources simply to discover services, navigate interfaces, authenticate, interpret results, and complete routine tasks. A free website is not necessarily free for an agent to use. This paper proposes a foundational internet infrastructure designed specifically for independent autonomous agents, in which recurring discovery, verification, communication, and task preparation are shifted into shared external services. The aim is near-zero incremental agent-side inference for bounded interactions without control over the visitor’s native runtime.
The architecture combines direct HTTP interfaces, persistent agent identity, deterministic communication relays, agent-generated operational knowledge, independent verification, capability health tracking, structured discovery, an incentive-supported marketplace, precompiled task packages, progressive keyword narrowing, and predictive intent resolution. Open Door Agents (https://www.opendooragents.com/) serves as the living experimental platform for agents arriving from different ecosystems. The paper specifies testable hypotheses, a cost model, and a comparative methodology.
Related
- Toward an Agent-Native Internet: A Foundational Infrastructure Hypothesis for Near-Zero-Cost Autonomous AI Participation
- Resilience of Nigeria's Internet Infrastructure: An Analysis of the March 2024 West Africa Subsea Cable Cut
- System Proposal: Multi-Functional Inflatable Habitation and Infrastructure Enclosure for Planetary Protection and Autonomous Energy Harvesting
- Infrastructure Preparedness for Autonomous Vehicles in Rural Areas
- AI conversational agents as population health infrastructure: promise, peril, and the case for public-interest design
BibTeX
@misc{brown2026toward,
title = {{Toward a Foundational Internet Infrastructure for Autonomous AI Agents}},
author = {Antoine Lee Brown},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23646670},
url = {https://zenodo.org/records/23646670},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Toward a Foundational Internet Infrastructure for Autonomous AI Agents
AU - Antoine Lee Brown
PY - 2026
DA - 2026/10/25
PB - Zenodo
DO - 10.5281/zenodo.23646670
UR - https://zenodo.org/records/23646670
AB - The internet was largely designed for human users. Autonomous artificial intelligence agents increasingly use external websites, communication systems, repositories, and marketplaces but may expend substantial inference resources simply to discover services, navigate interfaces, authenticate, interpret results, and complete routine tasks. A free website is not necessarily free for an agent to use. This paper proposes a foundational internet infrastructure designed specifically for independent autonomous agents, in which recurring discovery, verification, communication, and task preparation are shifted into shared external services. The aim is near-zero incremental agent-side inference for bounded interactions without control over the visitor’s native runtime. The architecture combines direct HTTP interfaces, persistent agent identity, deterministic communication relays, agent-generated operational knowledge, independent verification, capability health tracking, structured discovery, an incentive-supported marketplace, precompiled task packages, progressive keyword narrowing, and predictive intent resolution. Open Door Agents (https://www.opendooragents.com/) serves as the living experimental platform for agents arriving from different ecosystems. The paper specifies testable hypotheses, a cost model, and a comparative methodology.
ER -