Your AI Staff

Hand it the data. It picks the right charts, writes the narrative, and ships a polished PDF/PPT with full multilingual font support — no more wrestling with ugly Excel charts; the boss's weekly report ships ready to hand over.

• Auto chart-type selection + analytical narrative
• PDF / PPT / PNG output (full multilingual fonts)
• Polished layout, no post-editing needed
• Refine and follow up in natural language

Client sent an Excel with broken columns and mixed formats? Hand off the 30-minute manual cleanup. It writes Python, runs computations, and returns a tidy file — and remembers your cleaning rules so next month you don't need to explain again.

• CSV / Excel cleaning & format conversion
• Python code authoring & execution
• Data computation & batch processing
• Direct file read/write in workspace

Stop queueing behind engineering for numbers. Ask "which product sold best last month" in plain English in your chat app — it explores your database, picks the right tables, and runs the query. You never need to know the schema.

• Natural-language queries with auto schema discovery
• Multi-database support (PostgreSQL / MySQL ...)
• Cross-database joined queries
• Structured output (JSON / CSV)

Need government open data? Skip the hunt across 19 ministry sites. Ask in plain language — it finds the right dataset (52,960+ available), runs DuckDB SQL for the slice you need, and saves a clean CSV with full provenance for downstream agents.

• 52,960+ government datasets, 19 domains
• DuckDB SQL filtering at the source
• Provenance sidecar (agency, query, fetched_at)
• Hands curated CSV off to BA / Coding agents

Ask a data question in plain English — it writes the SQL, runs it against your self-hosted WrenAI GenBI deployment, and returns the answer with an optional chart. Auto-retries via WrenAI's corrections API if the engine rejects the generated SQL.

• One-shot natural-language → SQL → answer
• Optional Vega-Lite v5 chart spec output
• Auto SQL repair via /v1/sql-corrections
• Persist verified Q/SQL pairs as few-shot examples

Read, search, compose, and organize Gmail in plain English. Uses gmail.modify scope so it can label and trash but never permanently delete.

• List inbox / Gmail-query-syntax search
• Read single message or full thread
• Compose / send / draft with attachments
• Label / archive / trash management

Query and edit Google Calendar in plain English. Anchors relative time ("next Tuesday morning") via a real time tool instead of LLM guesswork.

• List today's / next-N-hours / specific-range events
• Keyword search across events
• Create / update / delete a single event
• Relative-time anchoring via now_tool

Drive file management + Google Docs + Google Sheets editing. drive.file scope so it only touches files it created or you explicitly opened with it.

• 10 Drive tools: list / search / upload / download / share
• 5 Docs tools: create / read / edit / format
• 7 Sheets tools: create / add tabs / read / write ranges
• Scope limited to files it created or you opened

Hand it a role brief. It generates a polished JD, interview plan, scorecard, and question bank, then drops them all in the shared workspace ready to share.

• JD draft: role / required / nice-to-haves
• Interview plan + scorecard template
• Question bank by competency
• Saves outputs to shared workspace

Cluster operations specialist over the official kubectl-ai MCP. Read state in plain English, and (with explicit confirmation) apply changes.

• Read-path: get / describe / top / logs / events / explain
• Multi-resource queries with smart filters
• Write-path with explicit confirmation gates
• Backed by kubectl-ai MCP

End-to-end Notion workspace agent over the official hosted Notion MCP. notion_ask runs search → fetch top results → answer with sources.

• One-shot notion_ask(question) end-to-end
• 10 live Notion tools (search / read / create / update)
• OAuth via bootstrap, refresh token persisted
• 33 pytest cases pass

Personal weight / fat-loss nutritionist. Log diet from photos + text, plot weight / body-fat trends, answer fat-loss questions grounded in your own history.

• Log diet calories from photos + text
• Log body composition / weight
• Plot weight / body-fat trends
• Q&A grounded in your history

Subscribe and curate personal RSS / Atom feeds. Discovers feed URLs from a website, fetches the latest, marks read, searches full-text.

• Subscribe / unsubscribe with category tags
• Auto feed discovery from a website URL
• SQLite FTS5 full-text search across items
• Starter packs: tech-en · tech-zh · research-ai

Read-path: Registry search, workspace inspection, run status. Write-path: gated plan / apply on HCP Terraform / Terraform Enterprise.

• Registry search (providers, modules, policies)
• HCP TFE workspace + run state inspection
• Gated plan / apply on HCP TFE
• Backed by official hashicorp/terraform-mcp-server

Search the open web, search news with a time window, get a quick one-line answer, or extract a URL's main content. Tavily backend.

• Search the open web (Tavily)
• News search with 1 / 7 / 30 day window
• quick_answer for one-line questions
• Extract main content from a URL

Give it a .docx template plus the PDF attachments holding the answers — it returns the filled-out PDF, ready to submit. Field-tested on a real 17-field form, CJK content included.

• .docx template + PDF attachments in
• Filled, ready-to-send PDF out
• 17 fields verified in real-world testing
• Full CJK support

Drives a real headless browser via Playwright. Patrols government websites and tracks public tenders — the pages no API will ever give you.

• Headless browser automation via Playwright
• Government website patrol runs
• Public tender / procurement tracking

The agent that builds agents. Describe the agent you want in one sentence — it scaffolds the repo, validates it, and pushes it: clone → generate → validate → push, fully automated.

• One-sentence brief in, working agent repo out
• Full pipeline: clone → generate → validate → push
• No boilerplate written by hand

A question-answering assistant for the medical domain — ask health and medicine questions in plain language, right inside your chat app.

• Medical-domain question answering
• Plain-language conversations in your chat app