Online broker Deriv has published Deriv One, a self-hosted enterprise AI platform, under an MIT licence. Its announcement, posted on 30 October with a 29 October dateline, describes a shared workspace, reviewed skills, agent registration and an app-building Lab. The public repository and licence confirm the software release.[1][2][3]
Deriv says 54 employees contributed 221 skills to the internal system. This is an operational inventory reported by the company, not an externally audited productivity measure. The platform began as a request for a skills registry and developed into a governed environment with access controls, secrets management, sandboxed execution and usage or cost tracking. Its catalogue is intended to be visible according to user role.[1]
Deriv has set a target to automate 75% of manual workflows by the end of 2026. It has not said that target is achieved, and the reported inventory of 221 internal skills is not a count of independently measured task replacements or confirmation that every internal skill is included in the public repository.[1]
What users receive — and what remains internal
The README describes a workspace, role-controlled skills, Slack agents, scheduled automations and sandboxed Python execution, connected to an adopter's own compatible model gateway. It distinguishes local development from production: production operators need a real identity proxy and must disable mock authentication. The company's production cloud plugins, internal integrations and some Lab features are excluded from the release.[1][2]
The MIT licence permits reuse and modification subject to retaining its notice, and supplies the software without warranty. The repository documents an architecture for governed AI use; neither publication nor the licence establishes that a new deployment meets a broker's security or compliance requirements.[2][3]
Published scope and limits
| Item | Details |
|---|---|
| Open-source licence | MIT |
| Internal skill inventory | 221 skills from 54 contributors, company-reported |
| Governance components | Role-based skills, agents, secrets, sandbox and usage tracking |
| Model access | Bring your own compatible gateway |
| Excluded from public release | Internal production plugins and some Lab features |
| Automation percentage | 75% by end of 2026 is a company target |
Sources & further reading
- Deriv Open-Sources Deriv One, Its Self-Hosted Enterprise AI Platform Deriv · 30 Oct 2026 · Accessed 4 Nov 2026
- Deriv One public repository and README Deriv · Accessed 4 Nov 2026
- Deriv One MIT licence Deriv · Accessed 4 Nov 2026