AI Coding Agents

Agency: UN Secretariat | Reference: RFIUNPD24626 | Type: Request for information Provision of AI Coding Agents for United Nations Global Operations 1. Purpose The United Nations is conducting a market survey to understand the availability, maturity and delivery models of AI Coding Agent solutions suitable for its worldwide operations. This is a market-research exercise only. It is not a solicitation, creates no obligation on the United Nations, and no costs incurred in responding will be reimbursed. Responding is not a condition for participating in any future solicitation. 2. Background The United Nations maintains a global operational footprint spanning Peacekeeping Missions, Special Political Missions, Offices Away from Headquarters, and Economic Commissions across all regions of the world. Its personnel operate in diverse environments, including headquarters locations, regional offices, field missions, and remote duty stations. To support digital transformation, software development, and operational innovation, the United Nations seeks to explore the adoption of Artificial Intelligence (AI) Coding Agent solutions. AI Coding Agents are software development assistants capable of supporting the full software development lifecycle through natural language interaction. These solutions can assist personnel in software engineering, scripting, automation, testing, debugging, documentation generation, code review, and application modernization activities. The participating vendors should be familiar with the work of the United Nations, and have the capability to offer the AI model inference services wherever United Nations has a presence. In line with the Sustainable Development Goals (SDGs) 9 and 12, this request for an Expression of Interest aims to: Promote innovation, digital transformation, and resilient technological infrastructure (SDG 9); Promote sustainable digital operations through efficient resource utilization, intelligent automation, and the reuse of technological assets (SDG 12). Indicative environment, for sizing purposes only: • Approximately 2,000 developers, engineers and data specialists • Users located across headquarters duty stations, regional offices, field missions and remote duty stations in all regions • Connectivity ranging from high-bandwidth headquarters networks to intermittent, bandwidth-constrained field links, with a limited number of restricted or air-gapped environments • Mixed code estate, spanning legacy systems and modern cloud-native applications      Specific Requirements/Information 1. Who may respond • Providers of AI models and/or AI coding agent products • Authorised resellers and licensing partners • Systems integrators and managed service providers • Consortia, provided one prime respondent is named 2. . Areas of interest Respondents may address one or more of the following. There is no expectation that any single supplier covers all of them. A. AI coding agent ecosystem — an agent (harness) supplied together with its underlying model(s), with a fixed monthly price with a defined request/token quota, and a path to consumption-based pricing once quota is exhausted. B. AI coding agent, harness only — model-agnostic, able to connect to third-party inference endpoints over documented open HTTP APIs. C. Coding plan, model only — fixed monthly price with a defined request/token quota, and a path to consumption-based pricing once quota is exhausted. D. Offline and hybrid capability — deployment in air-gapped or restricted-network environments, and hybrid inference combining a local model with a larger hosted model. E. Enabling services — deployment, integration, training, onboarding and global technical support. 3. Capabilities of interest The United Nations is interested in solutions offering, in whole or in part: Agent capability — decomposition of complex tasks into executable steps; semi-autonomous execution across design, coding, testing and refactoring under human oversight; approval gates for code changes and command execution; configurable safeguards against unauthorised actions; multi-file and multi-repository operation; context-awareness over large codebases; sandboxed execution; sub-agents; tool calling; automated documentation generation; model routing by task complexity; and support for open interoperability protocols. Model capability — code generation across multiple languages; code explanation and analysis; refactoring and optimisation; test generation; code review; debugging and root-cause analysis; architecture and design recommendations; database query generation; and Infrastructure-as-Code support. Administration — role-based access control; enterprise SSO; audit logging and activity tracking; quota management by user and organisational unit; and a management API. Governance and security — data protection and privacy controls; data processing declaration with any third party vendors; commitment of not using client data for training AI models, a zero-retention option; recognised information security certi CONTACT: • Email: tazegul.taganova@un.org