deep-research

Generates cited research reports from multi-source web searches using firecrawl and exa MCPs.

1|Updated Oct 11, 2025
One-click install
npx skills add https://github.com/ibytechaos/claude --skill deep-research-ibytechaos
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: deep-research
Source: https://github.com/ibytechaos/claude/tree/main/plugins/everything-claude-code/skills/deep-research
Command: npx skills add https://github.com/ibytechaos/claude --skill deep-research-ibytechaos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Answering complex questions thoroughly requires searching many sources, reading them in depth, and synthesizing findings with citations, which is slow and error-prone when done manually. ## Core Features & Use Cases - Multi-Source Search: Runs structured sub-question searches through firecrawl and exa MCP tools, gathering 15-30 unique sources per topic. - Cited Report Generation: Produces a structured markdown report with executive summary, themed sections, inline citations, source list, and methodology notes. - Parallel Research: Splits broad topics across subagents so multiple research threads run concurrently and merge into one report. - Use Case: Ask for a deep dive into the competitive landscape of AI code editors and receive a sourced report covering market players, trends, and key takeaways. ## Quick Start Research the current state of nuclear fusion energy and give me a cited report with key takeaways.

Frequently Asked Questions about deep-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run deep research with citations using Claude?

Ask Claude to research a topic in depth, and the skill breaks it into sub-questions, searches via firecrawl or exa MCP tools, reads key sources, and delivers a structured report with inline citations and a source list.

What MCP servers are required for web research?

At least one of firecrawl or exa must be configured in ~/.claude.json or ~/.codex/config.toml. Firecrawl provides search, scrape, and crawl tools; exa provides web search and crawling. Using both together gives the best coverage.

Can I research multiple sub-topics in parallel?

Yes, for broad topics the skill uses Claude Code's Task tool to launch parallel research agents, each handling different sub-questions. The main session then synthesizes their findings into the final report.

How does the research report handle unverified claims?

Every claim requires a source, and findings supported by only one source are flagged as unverified. If information cannot be found for a sub-question, the report explicitly states insufficient data rather than guessing.

What are the limitations of MCP-based web research?

Results depend on the configured MCP servers and their API access, so paywalled or blocked pages may be unreachable. Recency also depends on search indexing, though the workflow prioritizes sources from the last 12 months.