Code Intelligence
Code & Database Intelligence
Not a generic chatbot. DevSpec has read your code, your schema, your git history, and every past decision your team has made. When it helps you think through a problem, it's working from your actual reality — not hypotheticals.
Why don't we verify webhook signatures inline? Feels wrong.
You and Tom decided this on 12 May — inline verification blocked the worker pool under load, so verification moved out-of-band. The trade-off is recorded; the constraint was p95 latency, not security. If the load profile has changed, it's worth revisiting — but that's the history.
— it holds the thread. Even the threads you dropped.
Generic AI gives generic answers
Every AI tool claims to understand code. Few have read yours. Without your repo structure, your schema, and the history of why things are the way they are, suggestions are plausible guesses — and your team spends half the session correcting them.
A new developer joins. Within a week they know why every architectural decision was made, who made it, and what the alternatives were — without bothering a single senior engineer.
How it works
Connect your sources once. DevSpec keeps its understanding current and compounds it with every conversation.
Connect your repos
GitHub OAuth, no clones, no agents on your servers. DevSpec indexes structure and symbols across seven languages — TypeScript, JavaScript, Python, Go, Rust, Java — and reindexes automatically on every push.
Connecting a GitHub repository and watching the code index buildIndexed two hours ago, three commits behind — it tells you.
Connect your database
Schema introspection keeps tables, relationships, indexes, and row-level security in view. Built-in advisors flag missing RLS policies and unindexed foreign keys before they become incidents.
Database schema view with security and performance advisorsThe missing index gets named before production names it for you.
Memory compounds
Every decision, trade-off, and 'why we didn't do it the other way' is captured as structured memory — twelve types, from conventions to risks. The longer you use DevSpec, the sharper its suggestions, plans, and Autopilot output get.
Project memories: decisions, conventions, and risks recorded across sessionsSession 50 knows things session 5 had to be told.
Docs that stay true
DevSpec generates a living knowledge base for your project — organised by feature area, written for both engineers and end users — and reconciles it nightly against what actually changed. Human edits are preserved, conflicts surfaced.
Generated project documentation organised by feature areaDocumentation that notices when it goes stale.
Intelligence you can interrogate
Semantic code search
Search by symbol, dependency, or meaning — not keyword grep. The AI cites actual files and line ranges when it answers.
Schema advisors
Security and performance advisors read your live schema: missing RLS, absent indexes, risky defaults — flagged in conversation, before production finds them.
Git history awareness
Recent commits, branch comparisons, and per-file history are part of the context. "What changed since Tuesday?" is a question, not an archaeology project.
Freshness tracking
Every repo shows when it was indexed and how far behind HEAD it is. Push webhooks trigger incremental reindexing so the AI is never reasoning about stale code.
Memory with provenance
Memories carry confidence, tags, and lineage. Supersede or retract them as decisions evolve — conflicting memories are detected and surfaced, not silently merged.
Two-audience docs
Each feature area gets technical docs for the team and a user guide for everyone else, generated from the same source of truth.
Convinced?
Set up your first project in under five minutes. Connect your repo, start a session, and see what your duck can do.
Start Your Project14-day free trial · 1,500 credits included