Deckhand
An AI engineering agent you drive from chat. Ask a real offshore / oil & gas question in plain language — Deckhand runs the actual calculation against the right standard and returns a deliverable that shows its work.
How it works
Chat → action → result — with enforcement below the model.
Ask
Send an engineering question in plain language — no syntax, no software jargon required.
Run
Deckhand picks the right method and standard and runs the real calculation on the digitalmodel engine.
Show its work
You get the number, the governing clause, and the assumption ledger — nothing hidden.
Deliver
Calculation notes and code changes arrive as pull requests you can review and merge.
Hard, auditable guardrails
The enforcement sits below the model, not in a prompt.
Two ways to run it
the capability surface — free
A Telegram community for engineers trying Deckhand on open, MIT-licensed engineering work.
- Reads, analyses, documentation drafts and reviews
- Changes arrive as pull requests to a public sandbox
- No force-push, history rewrite, or deletes
- Scoped to open repos only — everything else returns 404
Private scopes — for client work
Confidential, proprietary, and licensed-solver work runs on isolated scopes with their own data, servers, and credentials.
- Dedicated repositories and fine-grained credentials
- Full reply visibility within your channel
- Domain-specific knowledge packs
- Nothing client-specific touches the open tier
What it covers
Drawn from ~30 engineering disciplines in the digitalmodel platform.
Bring a real question
Every capability is free to explore, and each one is traced to the standard it implements.
Explore capabilities →