OpenCode breach leaves users facing stolen API keys and unexpected bills

Expect time spent rotating API keys and auditing AI billing after OpenCode's platform breach; small teams face added risk until the incident is resolved.

What actually changed

OpenCode — described in coverage as an "open source AI coding agent" — had part of its online platform compromised. Attackers stole some users' API keys and used them to run up charges on those users' AI accounts. When the compromised agent sought help from other AI tools, those tools refused to inspect the malicious code, leaving the attacker's payload and its effects visible in operational logs.

The immediate technical change is not a feature update or pricing move but an operational security incident affecting credential handling and downstream billing tied to AI-provider accounts.

Who it affects

• Current OpenCode users whose API keys were stored or used via the platform; they already experienced unauthorised charges.
• Teams evaluating OpenCode as an alternative to Claude Code or Cursor, because the incident bears directly on platform security and key management.
• Any group that reuses or shares API keys across projects or team members, since the theft demonstrates that keys accessible through a platform can be exfiltrated and used to bill accounts.

What it costs or what it replaces

• Pricing: the announcement does not state pricing for OpenCode.
• Replacement: OpenCode is presented as a popular alternative to Claude Code and Cursor; organisations considering those tools could view OpenCode as a functional substitute. The incident, however, adds a cost in operational risk: affected users incurred unexpected charges on their AI accounts.

Because the report documents stolen keys and incurred charges, the non-financial cost — hours to investigate, rotate credentials, and reconcile billing — is concrete and likely immediate for affected teams.

What we don't know

• How many users were affected and the total financial impact.
• Which AI-provider accounts were charged and whether those providers will refund affected customers.
• The exact vulnerability vector in OpenCode's online platform and whether it has been fully patched.
• Whether OpenCode will offer reimbursement, credit or other remediation to impacted users.
• How OpenCode stored API keys (client-side, server-side, encrypted vault) and what key-rotation guidance it will publish.

What to do next

1. If you use OpenCode: rotate any API keys that were used with the platform, check recent AI-provider billing for unexpected charges, and revoke any exposed credentials. Treat keys used with OpenCode as compromised until told otherwise.
2. Pause any planned migration to OpenCode from Claude Code or Cursor until OpenCode publishes a post-incident report detailing the vulnerability, scope and remediation steps. Reconsider key management approaches that rely on shared long-lived credentials.
3. Adopt least-privilege API keys and short-lived credentials where possible, and log all AI-provider API usage so you can detect anomalous billing quickly.