Back to Gap Map Product / UX / Agent identity & authorization

OxDeAI

OxDeAI
open source / Overall score: 1.8

OxDeAI is an execution-time authorization library for AI agents. A deterministic policy engine decides each proposed action and, when it allows it, issues a signed single-use authorization bound to that exact action, which an enforcement gateway checks before the tool runs. The TypeScript packages also keep a hash-chained audit log and support single-hop delegation with narrower scope. Ange Yobo maintains it.

Openness

5 high confidence
5.0
license
Apache-2.0 (decision core, enforcement gateway, SDK and framework adapters)
source
public
core features withheld
no — the 884-path tree at b58468b carries the root Apache-2.0 LICENSE and Apache-2.0 LICENSE files in each package, with no ee/, enterprise/, commercial/ or proprietary/ path; the project is solo-maintained and sells no edition

OxDeAI's decision engine, enforcement gateway and adapters are all Apache-2.0 code in one public repository, and its demo runs entirely on a local machine with no outside service. A single maintainer runs the project, and no company sells a paid edition beside it.

Adoption

1 medium confidence
1.0

Adoption is read from npm downloads of @oxdeai/core and @oxdeai/guard, the decision and enforcement packages. The guard depends on the core, so one install can count in both, and downloads cannot show how many deployments put the gateway in front of a real tool.

Capability

3 medium confidence
3.0

OxDeAI decides each action before it runs, hands narrower authority to a second agent and keeps a hash-chained record of its decisions. Agents are named only by an identifier in the request rather than holding a credential of their own, and nothing waits for a person to approve an action.

Verified 2026-11-20, except some scores, still awaiting re-confirmation