Grant Scanner
Tell the agent what you fund or pursue. It opens SAM.gov and Grants.gov, runs the searches, follows listings into detail pages, and returns a comparison table sorted by deadline.
Learn moreBulleBrowser is a desktop browser with a built-in AI agent. Tell it what you need — a grant scan, an RFP comparison, a compliance review — and watch it open tabs, read pages, and return a structured answer.
Bring your own Anthropic API key. Your prompts go directly to your provider — never through us.
Preset skills for the most common research tasks at nonprofits, government agencies, and the firms that serve them.
Tell the agent what you fund or pursue. It opens SAM.gov and Grants.gov, runs the searches, follows listings into detail pages, and returns a comparison table sorted by deadline.
Learn morePaste 2 to 4 RFP URLs. The agent reads each end-to-end, lifts the deadline, scope, eligibility, value, and evaluation criteria, and hands back a side-by-side breakdown ready for a go / no-go memo.
Learn moreDrop a document. The agent reads every clause, flags issues against EEO, FERPA, and ADA — plus your own checklist — and returns quoted clauses with section references for fast partner review.
Learn moreBulle Consulting works with community colleges, nonprofits, and public agencies on grant strategy, RFP review, and EEO compliance — and a third of that work used to live in browser tabs. So we rebuilt the browser around the agent loop. The research half disappears. The strategy half is what your day is actually spent on.
Launch BulleBrowser, choose Grant Scanner, RFP Comparator, or Compliance Review from the agent panel — or describe the task in plain English.
The agent drives real sites with Chrome DevTools Protocol — clicking, scrolling, reading pages — and streams every step into the side panel so you can follow along.
A comparison table. A clause-by-clause memo. A sorted list of opportunities. Copy it into your deck or doc and ship the deliverable.
No telemetry, no routing through Bulle, no surprises in the agent loop.
Your API key lives encrypted in the OS keychain. Prompts and responses move between your machine and Anthropic directly — we never see them.
History, conversations, and saved skills live on your disk. There is no telemetry in v1, and your research stays where you can see it.
Destructive actions require explicit confirmation. A hard 25-step cap per task keeps agent runs predictable, traceable, and cheap.
A real Chromium browser with an agent loop wired to the active tab via the Chrome DevTools Protocol — deterministic, auditable, open source.
Each tab is an Electron WebContentsView attached to a single BaseWindow. The renderer is itself a separate WebContentsView that hosts the React chrome around the active tab.
The agent calls the Anthropic Messages API with a tool registry. Tool-use blocks are executed against the active tab via Chrome DevTools Protocol — navigate, click, type, read, extract — then results are appended and the loop continues.
contextIsolation on, nodeIntegration off, sandboxed renderer. The UI talks to the main process only through a narrow, typed IPC bridge. Your Anthropic key lives encrypted in the OS keychain.
Bring your own Anthropic API key. Your prompts stay yours. Open source on GitHub.
Version 0.1.3 Linux Preview · built & signed via GitHub Actions.