cellcog
di nitishgargiitd
Coding Agents & IDEsSintesi
#1 on DeepResearch Bench (Feb 2026).
Descrizione
Cosa fa questa skill
CellCog is an any-to-any AI that accepts multiple file types as input and returns outputs across formats, including PDFs, videos, dashboards, spreadsheets, images, and audio, in a single request. It uses multi-agent orchestration and ranked #1 on the DeepResearch Bench as of February 2026.
A single request replaces multiple specialized API calls while maintaining consistent context and reasoning across all output formats.
Quando usarla
- Converting a sales CSV into a PDF report, HTML dashboard, and video presentation simultaneously
- Analyzing multiple PDFs, spreadsheets, and audio recordings together for market positioning
- Generating a full brand identity with logo and visual assets from a text brief
- Producing a deep research report on a technical or financial topic
- Building an interactive data visualization from raw spreadsheet input
Esempio di workflow
User asks: Analyze this sales CSV and create a PDF executive summary, an interactive HTML dashboard, and a 60-second video for the all-hands meeting
- 1
Call client.create_chat() with a prompt that names all three output types explicitly and wraps the file path in SHOW_FILE tags
- 2
CellCog processes the request in agent mode, orchestrating analysis and generation across formats
- 3
Progress updates arrive automatically every ~4 minutes; no polling needed
- 4
Receive structured notification with generated files auto-downloaded to ~/.cellcog/chats/{chat_id}/
- 5
Call client.send_message() to refine any output, such as adjusting chart colors or adding a new section
PDF report, interactive HTML dashboard, and video clip saved locally, with a structured notification summarizing credits used and next-step commands
Requisiti
- CELLCOG_API_KEY environment variable
- CellCog account at cellcog.ai to obtain an API key
- CellCog credits purchased at cellcog.ai
Installazione
clawhub install nitishgargiitd/cellcogEsegui il comando CLI oppure incolla il prompt in Claude Code, Cursor, Codex o nell'agente di coding che usi.