- Train an agent: teach an agent your data and business rules, then review every change before it goes live.
- Build a dashboard with parameterized queries: ask for a dashboard with filters, add more filters in chat, and reuse the queries behind them.
- Build a dashboard with state and saved data: build a dashboard that keeps what people add, such as reviews and ratings, and shares it with everyone who opens it.
- Create a scheduled task for recurring analysis: run a prompt on a schedule, so a monthly digest or weekly check is ready without anyone asking.
- Run an automated analysis from an external trigger: give an alerting tool or script a URL; each event starts an investigation.
- Test an agent with evals: write test questions with expected answers and see why each one passed or failed.
- Analyze your agents’ usage in chat: ask about runs, failures, feedback and cost, and get charts and instruction fixes back.
- Use saved queries for faster, more reliable answers: save a reviewed query once, and the agent reuses it instead of writing new SQL.
- Route questions to the right agent: use one instruction and clear agent descriptions so each question reaches the right agent.
- Analyze documents with Lists: turn a folder of contracts into a typed table with a source quote for every value.
- Use BOW from Claude or another MCP client: connect Claude Code, Cursor or Claude on the web and ask about your data from your assistant.
- Git workflow: manage instructions as code with Git and CI/CD.
- Embed BOW Chat: add a BOW report and chat experience to your own product.
Guides
Guides
Step-by-step walkthroughs for common tasks in Bag of Words.
Each guide walks through one task from start to finish, with example prompts and screenshots from a real run on the Music Store demo agent.