Perform AI-powered research on any topic using Poe models with built-in web search via mcporter. Use this skill when you need a comprehensive, synthesized answer to a research question — not just search results, but actual analysis and reasoning. Triggers on: "research this topic", "deep dive into", "comprehensive analysis of", "investigate", "what's the state of", "compare X and Y", "give me a thorough overview", "I need to understand", "write a research brief on". This is different from web search (tavily_search) — it uses AI models that can reason, synthesize multiple sources, and provide structured analysis.
The poe-research MCP server provides AI-powered research through Poe models with built-in web search. Unlike regular search which returns links and snippets, this returns synthesized, reasoned answers.
Access it through mcporter: mcporter call poe-research.<tool> key=value.
| Tool | Purpose | When to use |
|---|---|---|
research | Quick research with a single query | Focused questions, quick answers |
deep_research | Multi-step deep research | Complex topics, thorough analysis |
mcporter call poe-research.<tool_name> key=value
Output is a comprehensive text response with sources.
Research a topic using Poe with built-in web search. Returns a comprehensive, sourced answer in a single pass.
mcporter call poe-research.research query="your research question"
| Parameter | Required | Default | Description |
|---|---|---|---|
query | ✅ | — | The research question or topic |
model | ❌ | GPT-5.4 | Poe model to use (e.g. GPT-5.4, Claude-Sonnet-4.6) |
reasoning | ❌ | — | Reasoning effort: low, medium, high (only for reasoning-capable models) |
Quick factual research:
mcporter call poe-research.research query="What are the key differences between SQLite and DuckDB for analytical workloads?"
With reasoning effort:
mcporter call poe-research.research query="Analyze the tradeoffs between microservices and monolith architecture for a 10-person startup" reasoning=high
Using a specific model:
mcporter call poe-research.research query="What's new in Python 3.14?" model="Claude-Sonnet-4.6"
Perform multi-step deep research on a topic. Runs an initial search, then follows up to synthesize and fill gaps. Slower but much more thorough.
mcporter call poe-research.deep_research topic="your topic"
| Parameter | Required | Default | Description |
|---|---|---|---|
topic | ✅ | — | The topic to deeply research |
model | ❌ | GPT-5.4 | Poe model to use |
Comprehensive topic analysis:
mcporter call poe-research.deep_research topic="State of WebAssembly in 2025: adoption trends, performance benchmarks, and practical use cases beyond the browser"
Technical comparison:
mcporter call poe-research.deep_research topic="Comparing Bun, Deno, and Node.js runtime performance, ecosystem maturity, and production readiness in 2025"
Investigation:
mcporter call poe-research.deep_research topic="How are major tech companies implementing RAG (Retrieval-Augmented Generation) in production? Architecture patterns, challenges, and lessons learned."
Need AI-powered research?
├── Focused question, quick answer → research
├── Broad topic, need depth → deep_research
├── Simple factual lookup → research
└── Complex multi-faceted topic → deep_research
research is faster (single pass). Good for focused questions where you need a synthesized answer.deep_research is slower (multi-pass). It follows up on gaps from the initial research. Good for complex topics where coverage matters more than speed.| Need | Tool | Why |
|---|---|---|
| Synthesized analysis of a topic | ai-research (this) | AI reasoning + web search combined |
| Quick web search for facts | tavily_search | Faster, raw search results |
| Deep multi-source research with full control | tavily_research | Structured research pipeline |
| Code examples from real repos | code-search | Actual code, not analysis |
| Library documentation | context7 or tavily_skill | Reference docs, not research |
AI research is the best choice when you need the AI to think about the answer — compare options, analyze tradeoffs, synthesize from multiple sources — not just retrieve information.