Stack-aware agent
The agent doesn't just retrieve from your docs. It reads your real stack — connected tools and the user's live context — before it answers.
What "stack-aware" means
Most support bots run a generic RAG pipeline: retrieve from your docs, answer. That breaks the moment a question depends on the user's actual account state ("why was I charged twice?", "why is my deploy failing?"). DuggAI investigates before it drafts: it can pull live data from the tools you connect, and it always has the user's context from the conversation.
Connected tools
From your dashboard → Connectors, you connect the apps your stack runs on. Once connected, the agent can read from them to ground an answer (and, with your approval, take an action). Available connectors:
- Stripe — subscriptions, invoices, charges.
- Sentry — recent errors and stack traces.
- GitHub — repos, issues, code.
- Supabase — your project data.
- Linear — issues and status.
- PostHog — product analytics.
- Vercel — deployments.
- Convex, Daytona, Autumn, Slack.
The agent only sees the connectors you've actually wired up. It checks what's connected, queries the relevant one for the user in front of it, and cites what it found.
User context
On the widget, the agent gets whatever you pass via identify and setContext— the signed-in user's id, email, traits, and the page they're on. On Discord, it has the channel and the thread. That context disambiguates questions a generic bot would get wrong.
Knowledge base
Alongside live tool data, the agent retrieves from your knowledge sources— your crawled docs and any markdown you upload — and from past resolved conversations. Tool data answers "what's true for this user right now"; the knowledge base answers "how does this product work."
What it won't do
If the agent doesn't have a source — no relevant doc, no connector that covers the question — it says so and escalates rather than inventing an answer. See How the agent decides.