Three steps and Bolt.new is building on Powabase.
Create a project at app.powabase.ai and open the Connect dialog (top-left in Studio). Copy the Project URL and the Service Role (Secret) Key. The Service Role key reaches everything: the agent and AI endpoints, plus full database access. Keep it on the server and never put it in the browser.
Set the Project URL and Service Role key in Bolt's server environment so the app can reach Powabase. Keep the key off the client.
BASE_URL=<Project URL>
API_KEY=<Service Role (Secret) Key>Tell Bolt what to build. It scaffolds and runs the app against Powabase. Pick an example below to start.
Real apps, each from one prompt. Copy any of them and give it to Bolt.new.
Ingest our product docs, help center, and resolved tickets into a knowledge base, embedding on upload. Expose a support agent that answers with inline citations, streams over SSE, and hands off to a human when confidence is low. Add email and Google sign-in, isolate every conversation per organization with row-level security, and ship it as an embeddable chat widget.
Index our wikis, runbooks, and exported threads into a knowledge base, chunking and embedding on upload. Expose an agent that answers with citations and respects per-department access, keep it fresh with scheduled re-indexing of changed sources, and add SSO sign-in plus an analytics view of unanswered questions.
Model tickets, organizations, and agents with row-level security, and auto-triage incoming tickets with an agent that predicts category and priority and drafts a reply from the knowledge base. Escalate low-confidence cases, run SLA timers as scheduled jobs, add email auth, and provide a live queue dashboard.
Model companies, contacts, deals, and activities in Postgres with row-level security per team and roles for reps and managers. Build a pipeline board with stage tracking, an activity timeline, and semantic search across notes, plus a dashboard of weighted pipeline and win rate. Wire OAuth sign-in and an audit log of every change.
Pull RSS feeds and news APIs on a schedule, deduplicate and embed each article, and run an agent that clusters related stories, summarizes them, and tags entities and sentiment. Let users define watchlists and alert rules, isolate data per workspace, and deliver a realtime feed plus a daily email digest via a cron job.
Accept PDF and image uploads to storage, then extract vendor, line items, totals, and dates with an extraction agent and write validated records to Postgres, routing low-confidence fields to human review. Add approval workflows, per-team access control, exports to accounting formats, and a searchable archive.
