Glean opens headless agent development inside Claude Code, Cursor, and Codex
Glean’s generally available headless agent building feature lets teams define, test, version, and publish enterprise agents as code while running them on the company’s permission-aware context layer.
AT A GLANCE
Key points
- Agents are defined in spec.yaml and instructions.md files
- Developers can test, evaluate, version, and roll back changes
- Published agents retain Glean permissions and audit trails
- Headless building currently supports auto mode agents
What is changing?
Glean has made headless agent building generally available, allowing developers to create enterprise AI agents from Claude Code, Cursor, Codex, or Git. Agent configuration lives in a spec.yaml file while behavior instructions live in instructions.md, so teams can review and version them alongside application code.
Developers can run agents with representative inputs, inspect results and evaluations, and roll back changes. A finished agent is published to Glean’s managed runtime rather than separate infrastructure, retaining the company knowledge graph, permissioned access across more than 275 enterprise sources, and an audit trail.
Glean says the current release supports auto mode agents, while workflow mode agents must still be built in the visual Agent Builder. The distinction shows how code-first development can work with enterprise governance, while not every agent type is yet handled through the same workflow.
Explore the original sources
Source date: 21 October 2026.
Published by AI Radar: 22 Oct 2026, 10:08 (Istanbul, UTC+3). Updated: 22 Oct 2026, 11:14.
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…