AgentField
AgentField builds open-source infrastructure for building, scaling, and governing autonomous AI agents. Its control plane turns agents written in Python, TypeScript, or Go into production services with orchestration, identity, policy enforcement, durable execution, memory, and auditability, running on models and infrastructure the customer controls.
At a Glance
- Developers and engineering teams building autonomous software
- Banks and financial-services companies
- Energy companies
- Healthcare organizations
- +5 more
AI Tools by AgentField
(2)
SWE-AF
Multi-Agent Software Engineering Runtime
Agentfield
AI Agent Deployment Control Plane
Discussions
No discussions yet
Be the first to start a discussion about AgentField
Latest News
SWE-AF repository reached commit #152; latest merge added issue-level implementation support across its planner and fast runtimes.
AgentField repository published v0.1.142-rc.9, continuing active development of the open-source AI backend/control plane.
AgentField published a tutorial series showing local/personal control planes, self-tuning rollouts, reactive agents, dynamically built pipelines, agent fleets, and long-running autonomous jobs.
AgentField published How an AI-Native Engineering Team Does Code Review and announced/open-sourced PR-AF, its risk-map-oriented multi-agent pull-request reviewer.
Products & Services
Apache-2.0 open-source backend/control plane for deploying agents as APIs or microservices. It provides routing, coordination, queues, retries, streaming, durable long-running execution, memory, observability, canary/A-B/blue-green deployment, cryptographic DIDs, policy enforcement, and signed verifiable audit receipts.
Open-source software factory and coding-agent product. It coordinates teams of coding agents from one window, decomposes work into a dependency-aware task hypergraph, assigns models per seat through Pareto crewing, supports specialists and managers, and can run jobs locally over SSH with cron, hooks, watchers, tests, and durable sessions.
Open-source autonomous software-engineering fleet/orchestration system. It coordinates coding harnesses such as Claude Code, Codex, OpenCode, and related agents across isolated worktrees, with planning, implementation, verification, retries, and draft-PR delivery.
Open-source AI-native pull-request reviewer that analyzes the risk dimensions introduced by a change, spawns parallel reviewers, challenges findings adversarially, and surfaces a risk map as GitHub comments rather than a simplistic pass/fail verdict.
Market Position
AgentField positions itself as an open-source infrastructure/control-plane layer rather than another prompt framework: the Kubernetes-style runtime handles deployment, routing, durable execution, scaling, and observability, while an Okta-like identity and policy layer gives every agent cryptographic identity, delegated authority, and verifiable audit evidence. This differentiates it from application-level agent frameworks such as LangGraph and from coding-specific or review-specific tools such as Claude Code, Codex, Cursor, CodeRabbit, Qodo, and Greptile; CodeAF, SWE-AF, and PR-AF are specialized products built on or alongside the same platform.
Leadership
Founders
Oktay Goktas, PhD
Co-founder and CEO. Previously co-built Agnostiq and its open-source distributed-computing platform Covalent; Agnostiq was acquired by DataRobot in 2025.
Santosh Kumar Radha, PhD
Co-founder and CTO. Previously co-built Agnostiq and Covalent with Oktay Goktas; Agnostiq was acquired by DataRobot in 2025.
Executive Team
Oktay Goktas, PhD
Co-founder and Chief Executive Officer
Previously co-built Agnostiq and Covalent; Agnostiq was acquired by DataRobot in 2025.
Santosh Kumar Radha, PhD
Co-founder and Chief Technology Officer
Previously co-built Agnostiq and Covalent with Goktas; Agnostiq was acquired by DataRobot in 2025.
Board of Directors
Founding Story
AgentField was created around the founders' belief that software is moving beyond interfaces toward autonomous systems that negotiate, act, and complete intent. Drawing on a decade of distributed-computing and infrastructure work, the founders set out to provide the missing foundation for safe, predictable, identity-bearing agents: a Kubernetes-style execution/control plane combined with cryptographic identity, enforceable permissions, and verifiable audit trails.
Business Model
Revenue Model
The core software is open source under Apache-2.0 and can be installed on a laptop or customer-owned infrastructure. AgentField's enterprise model is based on private deployment, design partnerships, organizational governance, named engineering support, and SLAs; enterprise customers run the stack in their own VPC, on-premises, or air-gapped environment rather than paying a per-agent cloud meter.
Target Markets
- Developers and engineering teams building autonomous software
- Banks and financial-services companies
- Energy companies
- Healthcare organizations
- Logistics and supply-chain operators
- Insurance companies
- Production multi-agent backends and autonomous software
- Financial risk, transaction, and claims-processing workflows
- Healthcare and patient-record workflows
- Supply-chain, logistics, and freight decision automation
- Energy and critical-infrastructure operations
- Enterprise software engineering and autonomous coding fleets
- JPMorgan
- ServiceNow
- National Institutes of Health
- SAP