Deep research powered by GREP — investigate APIs, libraries, companies, markets, and any topic with sourced citations. Build interactive HTML apps, slidedecks, and spreadsheets backed by research. Route to one of 27 domain experts (legal, medical, patent, financial, real estate, supply chain, maritime, etc.). Use when you need to research something, fact-check a claim, plan an implementation with real docs, create a new skill, build a deliverable, install Grep as an MCP server, check GREP account status, or upgrade your plan. Trigger on: "research", "look up", "investigate", "fact check", "what's the best way to", "how does X work", "find out about", "build me a", "make a slidedeck/spreadsheet", "use the X expert", "follow up on", "research using these files", "install grep MCP", "GREP", or any task where knowing current facts would improve the outcome. Use proactively before writing code against unfamiliar APIs.
Authenticate with GREP deep research. Use when the user needs to log in to GREP, authenticate, set up GREP credentials, or when a GREP API call returns 401 unauthorized. Supports direct arguments for agent use — /grep-login EMAIL to login with OTP, /grep-login --api-key KEY to authenticate with an API key.
Attach a Grep MCP server to .mcp.json so Claude has 4 native research tools (research_create, research_get, research_files_list, research_file_read) without writing curl. Use when the user wants Grep available as MCP tools, says "install Grep MCP", "set up Grep MCP server", or wants to research via MCP instead of HTTP. Reads the existing .mcp.json (or creates one), merges the grep entry, and verifies the server responds to tools/list.
Create a high-quality SKILL.md for any AI agent skill, powered by GREP deep research. Use when the user wants to build a new skill, plugin, or tool integration for Claude Code, Cowork, or OpenClaw. Also use when the user says "create a skill", "make a skill for X", "turn this into a skill", "build a plugin", or when they've been working through a workflow and want to capture it as a reusable skill. Researches the target domain first, then generates a production-ready skill file.
Deep research via GREP (typically ~5 minutes) with sourced citations. Use for most research tasks — investigating APIs/libraries before writing code, company research, market analysis, verifying complex claims, comprehensive docs lookups, fact-checking with sources. ALSO use proactively when planning to write code against unfamiliar APIs, SDKs, protocols, or libraries — sourced research beats guessing from memory. Use /quick-research (~25s) for simple lookups, or /ultra-research (up to 1 hour) for exhaustive investigations.
Build interactive HTML web apps via Grep — websites, dashboards, data explorers, calculators, comparison tools, charts, custom UIs. Use when the user asks for an interactive app, "build me a tool that does X", "make an interactive Y", or anything requiring a runnable HTML/JS deliverable beyond a static report. Routes to the app-builder expert with effort=build (~$2.00, 10-15 min). Returns an index.html in the job workspace; the skill prints the file URL when ready.
Multi-step research workflow — chains a quick orientation pass, then a deep investigation, then optional ultra-research, then an optional build artifact (slidedeck / app / spreadsheet). Use when the user wants thorough end-to-end research with a deliverable, says "do a full investigation on X", "research and then make me a presentation about X", or "comprehensive analysis of Y with a writeup". Threads job_ids through reference_jobs so each step builds on the prior step's findings without re-running them.
Research using user-supplied files as context — uploads PDFs, CSVs, text files, or images to Grep before submitting the research job, so the agent can analyse the user's documents alongside web research. Use when the user attaches files to the conversation, says "research using these documents", "fact-check this PDF", "summarise this report and find related work", or "use this CSV as input for research".