compliance
Can I use AI coding agents on NDA client code?
A contractor’s checklist for using Claude Code-style agents on code covered by a client NDA.
Most NDAs do not ban tools; they restrict disclosure of confidential information to third parties and require you to protect it. Sending source code to an AI provider is a disclosure to a third party, so what matters is what that party does with it: whether it stores prompts, trains on them, and who can read them.
A zero-retention backend narrows that disclosure to transient processing: the code is processed to produce a reply and not kept. You still need to tell the client if the NDA or the master agreement requires it, and some clients forbid any external processing at all.
Checklist
- □Read the NDA and MSA for “third party”, “subprocessor” and “AI” clauses.
- □Ask the client in writing if AI coding tools are allowed, and which.
- □Use a backend that stores no prompts or code and does not train on them.
- □Keep secrets (.env, keys) out of the agent’s context with ignore files.
- □Attach the provider’s security summary to your answer.
This is practical guidance, not legal advice. Check your contracts with the client or a lawyer.
What Hushwire gives you
A backend for coding agents that stores no prompts, code or completions, never routes to OpenAI or Anthropic, and gives you a downloadable security summary to attach to the client’s questionnaire.
Try it before you configure anything
The playground runs the same open coding models. No sign-up, 20 messages a day, and nothing you type is stored.