Ollama and Continue.dev handle your inline coding privately. Hey Eduardo handles everything else — NDAs, security findings, client emails, architectural docs, research with sensitive context. All on-device. Nothing leaks.
The inline-coding-AI problem (Copilot, Cursor) is well-known. The chat-and-document AI problem is bigger and barely discussed — and it's where most of the legally-sensitive material actually ends up.
Since April 24, 2026, GitHub trains on Copilot Free, Pro, and Pro+ interaction data unless you opt out. Cursor had seven CVEs in 2025. The well-known answer: Ollama + Continue.dev or similar local-first IDE setup.
The NDA you paste into ChatGPT. The security audit you discuss with Claude. The client email you have Gemini help draft. This is where Hey Eduardo fits — on-device chat AI for everything outside the editor.
Local coding model in your IDE + Hey Eduardo for chat/research/documents = end-to-end privacy. Your client material never reaches any vendor, regardless of which AI you reach for.
Reading and reasoning about confidential material. Client NDAs. MSAs. Security audit findings. Architectural design docs. Incident postmortems. Sensitive Slack threads. Internal compliance documents. All paste-and-discuss workflows where the source material can't leave your Mac.
Research with sensitive context. Working through a problem that involves client business logic or data structures, without leaking the problem context to a third party.
Drafting communications. Client emails, status updates, apologies, contract responses — drafted with the full context of the situation without that context being logged anywhere off-device.
Before you sign a client NDA, check what it asks of you after the project ends.
Select the agreement, press Option-Shift-A, and ask.
A year without working directly for their clients, 10 days to return everything, and 3 years of confidentiality.
Then ask about AI tools.
Not without the company's written consent.
Section 2.3 bars hosted AI tools without written consent.
Paste this NDA or the client's code into ChatGPT or Claude, and it goes to another company's servers.
In local mode, Hey Eduardo reads it right here on your Mac. Nothing is sent anywhere.
Hey Eduardo. heyeduardo.app
No. Hey Eduardo is for the chat/document tier — reading NDAs, analyzing audit findings, drafting client emails, working through architectural decisions. For inline coding, pair it with Ollama + Continue.dev for a fully local stack.
In local mode it runs entirely on your Mac. Every NDA you paste, every requirement doc you analyze, every confidential design discussion — none of it leaves your hardware. ChatGPT logs everything and is under a federal preservation order.
Most NDAs prohibit disclosing client material to third parties. Submitting client material to consumer AI tools may be a contract breach. Local AI removes the third party entirely.
Since April 24, 2026, GitHub uses interaction data from Copilot Free, Pro, and Pro+ for model training unless you opt out in your Copilot settings. This is exactly the unilateral vendor change local AI protects you from. Your hardware, your terms.
One-time purchase from $49. 14-day money-back guarantee. Apple Silicon Macs only.