JPMorganChase widens Claude Code use beyond engineers to product, risk and wealth teams
Anthropic says agents run inside the bank's Devspace platform without direct access to employee credentials, and an enforced pipeline gate means Claude cannot approve its own changes.
Why it matters
The largest US bank is putting a coding agent in front of staff who are not engineers, which makes it one of the clearest tests of agentic AI inside a heavily regulated firm. The controls matter as much as the rollout: agents work on the bank's own infrastructure, each new Claude model passes internal approval, and people stay accountable for every production change. Watch for other banks copying the setup, and for any sign of how much engineering capacity JPMorganChase says it has freed.
Key facts
- Announced by Anthropic on 9 November 2026
- Claude Code runs inside Devspace, the bank's containerised platform
- Agents get no direct access to employee credentials
- New Claude models pass the bank's own approval process
- Teams in product, risk and controls, and Asset & Wealth Management