Perplexity Research History MCP Server

Multi-model synthesis for comprehensive research.

MANDATORY WORKFLOW - DO NOT SKIP:

Step 1: Check Existing Research

Use search_research() first to check for existing research on this topic.

Step 2: Present Research Plan

BEFORE calling this tool, present a plan explaining:

• Why multi-model synthesis is warranted
• Which pattern you'll use and why
• Expected time (30-120 seconds for deep research)
• What each model phase will investigate

Step 3: Wait for Approval

DO NOT call until user approves. Multi-model queries are expensive and time-consuming.

Step 4: Execute with Plan

After approval, call with approved_plan parameter.

PATTERNS:

• fact-reasoning: Factual research (sonar-pro) → Reasoning analysis (sonar-reasoning-pro) Use when: Need verified facts + causal explanation
• quick-deep: Quick overview (sonar) → Deep research (sonar-deep-research) Use when: Initial assessment then comprehensive dive
• truthtracer: Full verification (sonar-pro → sonar-reasoning-pro → sonar-deep-research) Use when: High-stakes decisions, fact-checking, due diligence
• multi-perspective: Multiple analytical angles Use when: Controversial topics, trade-off analysis

OUTPUT STRUCTURE: Results will include:

• Summary of key findings
• Findings by approach (what each model found)
• Areas of agreement (high confidence)
• Areas of conflict (flagged for user)
• Confidence assessment
• Combined citations

Results saved with raw JSON backup:

• Thread file: ~/Documents/Perplexity Research/[topic].md
• Raw response: ~/Documents/Perplexity Research/raw/[timestamp]_[topic].json

Show available Perplexity models and when to use each.

Provides guidance on model selection for different query types.

List all saved research threads with dates and summaries.

Shows what research has been done previously.

Read the full content of a past research thread.

Use after list_research_threads to read specific research.

Intelligent Perplexity search with auto-logging.

MANDATORY WORKFLOW - DO NOT SKIP THESE STEPS:

Step 1: Check Existing Research

BEFORE calling this tool, use search_research() to check for existing research on this topic. Build on prior findings and avoid duplicate queries.

Step 2: Present Research Plan

BEFORE calling this tool, present a Research Plan to the user:

## Research Plan

Objective: [Restate the specific question being answered in precise terms]
Scope: [What's in] | [What's explicitly out]
Methodology: [Model selected] because [rationale] | Sources: [prioritization]
Expected Output: [Brief answer / Detailed report / Comparative analysis]

### Alternatives to Consider
- [Alternative framing 1] — might be better if [condition]
- [Alternative framing 2] — worth considering because [reason]

### Questions Before Proceeding
- [Clarifying question if ambiguity exists]
- [Assumption being made that user might want to challenge]

### Potential Blind Spots
- [What this approach might miss]
- [Bias in source prioritization]

Awaiting your approval before proceeding.

Step 3: Wait for Approval

DO NOT call this tool until user explicitly approves the plan.

Step 4: Execute with Plan

After approval, call this tool with the approved_plan parameter containing your plan.

Step 5: For Follow-ups

For follow-up queries within established scope, use brief checkpoint: "Follow-up within established scope: [sub-question] using same methodology. Proceed?"

MODEL SELECTION (if not specified, will auto-select):

• Simple factual lookup (what is X?) → sonar
• Analysis, reasoning, most queries → sonar-reasoning-pro (default)
• Exhaustive research, comprehensive reports → sonar-deep-research
• RCA/debugging → sonar-reasoning-pro (mandatory for causal reasoning)

WHEN PRESENTING RESULTS, include:

• Confidence level (High/Medium/Low) based on source agreement
• Note any conflicts between sources
• Distinguish verified facts from inferences

Results saved with raw JSON backup:

• Thread file: ~/Documents/Perplexity Research/[topic].md
• Raw response: ~/Documents/Perplexity Research/raw/[timestamp]_[topic].json

Search past research for existing findings.

CALL THIS BEFORE making new queries to avoid duplicate work and build on prior findings. Searches across all thread files in ~/Documents/Perplexity Research/ (excludes raw/ directory).