Agent workspace
Each card opens a self-contained tool. More features will appear here over time.
Schedules, logs & AI briefings
Weekly recurring tasks, completion logs with optional ratings, voice assistant with page context, and agents for monitor alerts and daily summaries. Includes Multi-Agent Assessment (Claude) from the Task Hub page. Bring your own OpenAI key or use server config.
Claude · concierge → committee
Demo form validated by multiple Claude agents in sequence—investigator, administrator, and a grading committee—with logs and a downloadable transcript. Uses ANTHROPIC_API_KEY on the server.
Three phrases → three roasts
Enter three words or phrases; each comedian must use all three in one joke—traditionalist, clown, and funniest-on-earth take turns. Optional speech. Uses your OpenAI key (BYOK) or the server’s.
Natural language → SQL → reports
Agents that turn plain-English requests into SQL, run them safely against your data, and return reports with charts using the best-fit visualization. (In development.)
Prompt → implementation spec
Multi-agent pipeline over curated rule examples and P21 training docs: NL understanding, doc-grounded T-SQL sketches, and synthesis into a spec you can implement in P21. Nothing is written to P21 automatically.
SOPs → guided runs for the floor
Supervisors upload JSON checklists; an agent expands each step into a walkthrough. Assign workers by email or SMS—they get a private link to complete steps and log notes. Built for warehouse, DC, and counter workflows.