OpenRouter Agents MCP Server

# OpenRouter Agents - Comprehensive Use Cases & Workflows ## 🎯 **Quick Start: Single-Prompt Workflows** For platforms supporting MCP prompts, use these single-command workflows: ### **Research Workflow Prompt** ```mcp research_workflow_prompt { "topic": "quantum computing applications in machine learning", "costBudget": "low", "async": "true" } ``` --- ## 📋 **Domain-Specific Use Cases** ### 1. **Technical Research & Analysis** #### **Domain**: System Architecture & Engineering #### **Problem**: Understanding complex distributed systems and implementation patterns #### **Research Strategy**: Multi-source technical analysis with authoritative citations ```xml <workflow> <step1 tool="search_web"> <params>{ "query": "microservices event-driven architecture patterns 2025", "maxResults": 5 }</params> <purpose>Gather current industry perspectives</purpose> </step1> <step2 tool="fetch_url"> <params>{ "url": "[top_authoritative_source]", "maxBytes": 50000 }</params> <purpose>Deep-dive into primary technical documentation</purpose> </step2> <step3 tool="conduct_research"> <params>{ "query": "Compare event-driven microservices vs traditional REST APIs: performance, scalability, complexity trade-offs", "textDocuments": ["fetched_content"], "costPreference": "low", "audienceLevel": "expert" }</params> <purpose>Comprehensive analysis with ensemble model consensus</purpose> </step3> <step4 tool="research_follow_up"> <params>{ "originalQuery": "microservices architecture comparison", "followUpQuestion": "What are the specific security considerations and DevOps implications?" }</params> <purpose>Targeted follow-up for operational concerns</purpose> </step4> </workflow> ``` **Expected Outcome**: Technical architecture report with 15-25 citations, consensus across multiple models, specific implementation guidance --- ### 2. **Market Intelligence & Competitive Analysis** #### **Domain**: Business Intelligence & Market Research #### **Problem**: Gathering competitive landscape and trend analysis #### **Research Strategy**: Async processing with knowledge base integration ```xml <workflow> <step1 tool="get_past_research"> <params>{ "query": "AI market competitive analysis", "limit": 5 }</params> <purpose>Leverage existing knowledge base</purpose> </step1> <step2 tool="submit_research"> <params>{ "query": "AI SaaS market competitive landscape Q3 2025: key players, market share, pricing strategies", "costPreference": "low", "outputFormat": "briefing", "audienceLevel": "intermediate" }</params> <purpose>Async comprehensive market research</purpose> </step2> <step3 tool="get_job_status" repeat="true"> <params>{ "job_id": "[returned_job_id]" }</params> <purpose>Monitor progress with event streaming</purpose> </step3> <step4 tool="rate_research_report"> <params>{ "reportId": "[final_report_id]", "rating": 5, "comment": "Excellent market analysis" }</params> <purpose>Quality feedback for continuous improvement</purpose> </step4> </workflow> ``` **Expected Outcome**: Market intelligence briefing with competitive positioning, trend analysis, and actionable insights --- ### 3. **Creative Strategy & Innovation Research** #### **Domain**: Design Thinking & Creative Solutions #### **Problem**: Developing innovative approaches and creative strategies #### **Research Strategy**: High-cost models for creative reasoning with multimodal analysis ```xml <workflow> <step1 tool="list_models"> <params>{ "refresh": false }</params> <purpose>Identify available creative and vision-capable models</purpose> </step1> <step2 tool="conduct_research"> <params>{ "query": "Innovative UX design patterns for AI interfaces: emerging trends and user psychology", "costPreference": "high", "audienceLevel": "expert", "outputFormat": "report" }</params> <purpose>High-quality creative analysis with sophisticated models</purpose> </step2> <step3 tool="search" parallel="true"> <params>{ "q": "UX design AI interfaces", "scope": "reports", "rerank": true }</params> <purpose>Cross-reference with existing research for synthesis</purpose> </step3> <step4 tool="research_follow_up"> <params>{ "originalQuery": "AI interface UX patterns", "followUpQuestion": "How do these patterns apply specifically to conversational AI and voice interfaces?" }</params> <purpose>Targeted refinement for specific implementation domains</purpose> </step4> </workflow> ``` **Expected Outcome**: Creative strategy recommendations with design principles, user psychology insights, and implementation guidelines --- ### 4. **Multimodal Research & Vision Analysis** #### **Domain**: Visual Data Analysis & Multimodal Intelligence #### **Problem**: Analyzing charts, diagrams, and visual content with contextual research #### **Research Strategy**: Vision-capable models with document integration ```xml <workflow> <step1 tool="conduct_research"> <params>{ "query": "Analyze the data visualization trends and explain statistical significance", "images": [{ "url": "data:image/png;base64,[chart_data]", "detail": "high" }], "costPreference": "low", "audienceLevel": "expert", "includeSources": true }</params> <purpose>Vision-assisted analysis with statistical interpretation</purpose> </step1> <step2 tool="search_web"> <params>{ "query": "data visualization best practices statistical significance 2025" }</params> <purpose>Gather authoritative sources on visualization principles</purpose> </step2> <step3 tool="fetch_url"> <params>{ "url": "[top_statistical_source]" }</params> <purpose>Deep-dive into statistical methodology</purpose> </step3> <step4 tool="conduct_research"> <params>{ "query": "Synthesize findings: chart analysis + statistical methodology + best practices", "textDocuments": ["statistical_methodology_content"], "costPreference": "low" }</params> <purpose>Comprehensive synthesis combining visual and textual analysis</purpose> </step4> </workflow> ``` **Expected Outcome**: Detailed visual analysis report with statistical validation and methodology recommendations --- ### 5. **Knowledge Base Management & Quality Assurance** #### **Domain**: Information Management & Research Quality #### **Problem**: Maintaining research quality and leveraging historical insights #### **Research Strategy**: Database-driven research with quality controls ```xml <workflow> <step1 tool="db_health"> <purpose>Verify system readiness and database integrity</purpose> </step1> <step2 tool="search"> <params>{ "q": "artificial intelligence ethics", "scope": "reports", "k": 10, "rerank": true }</params> <purpose>Semantic search of existing knowledge base</purpose> </step2> <step3 tool="get_past_research"> <params>{ "query": "AI ethics frameworks", "limit": 5, "minSimilarity": 0.7 }</params> <purpose>Retrieve relevant past research for context</purpose> </step3> <step4 tool="conduct_research"> <params>{ "query": "AI ethics frameworks: current standards, regulatory developments, implementation challenges", "costPreference": "low", "audienceLevel": "intermediate" }</params> <purpose>New research building on existing knowledge</purpose> </step4> <step5 tool="backup_db"> <params>{ "destinationDir": "./backups" }</params> <purpose>Preserve research artifacts</purpose> </step5> </workflow> ``` **Expected Outcome**: Well-grounded research report leveraging institutional knowledge with data integrity assurance --- ### 6. **Cost-Optimized High-Volume Research** #### **Domain**: Efficient Research Operations #### **Problem**: Processing multiple research queries with minimal cost #### **Research Strategy**: Intelligent model routing with caching optimization ```xml <workflow> <step1 tool="list_models"> <params>{ "refresh": true }</params> <purpose>Get current model availability and pricing</purpose> </step1> <step2 tool="query" iterative="true"> <params>{ "sql": "SELECT query, report_id FROM research_reports WHERE created_at > NOW() - INTERVAL '7 days' ORDER BY created_at DESC LIMIT 10" }</params> <purpose>Identify recent research patterns for cache optimization</purpose> </step2> <step3 tool="submit_research" batch="true"> <params>{ "query": "[bulk_research_topics]", "costPreference": "low", "outputFormat": "bullet_points" }</params> <purpose>Async batch processing with cost-effective models</purpose> </step3> <step4 tool="reindex_vectors"> <purpose>Optimize search performance for future queries</purpose> </step4> </workflow> ``` **Expected Outcome**: High-volume research processing with 60-80% cost savings through intelligent caching and model selection --- ## 🛠 **Advanced Integration Patterns** ### **Pattern A: Iterative Deep-Dive Research** ```mcp # Step 1: Initial broad research conduct_research { "query": "blockchain scalability solutions", "costPreference": "low" } # Step 2: Follow-up on specific findings research_follow_up { "originalQuery": "blockchain scalability", "followUpQuestion": "How does sharding compare to layer-2 solutions?" } # Step 3: Cross-reference with knowledge base search { "q": "blockchain sharding layer-2", "scope": "reports", "rerank": true } # Step 4: Final synthesis conduct_research { "query": "Synthesize comprehensive blockchain scalability analysis", "textDocuments": ["previous_research"] } ``` ### **Pattern B: Multimodal Evidence Synthesis** ```mcp # Step 1: Vision analysis conduct_research { "query": "Analyze market data charts and explain trends", "images": [{"url": "data:image/png;base64,...", "detail": "high"}], "costPreference": "low" } # Step 2: Document integration conduct_research { "query": "Correlate visual analysis with financial reports", "textDocuments": [{"name": "q3_report.pdf", "content": "..."}], "structuredData": [{"name": "metrics.csv", "type": "csv", "content": "..."}] } ``` ### **Pattern C: Quality-Controlled Research Pipeline** ```mcp # Step 1: Health check db_health # Step 2: Research execution submit_research { "query": "AI safety alignment research 2025", "costPreference": "high" } # Step 3: Quality assessment rate_research_report { "reportId": "[report_id]", "rating": 4, "comment": "Good coverage, needs more recent sources" } # Step 4: Knowledge preservation backup_db { "destinationDir": "./backups" } ``` --- ## 📊 **Cost Optimization Guidelines** ### **Model Selection Matrix** | Query Type | Recommended Models | Cost/Token | Use Case | |------------|-------------------|------------|----------| | **Simple Factual** | `deepseek/deepseek-chat-v3.1` | $0.0000002 | Quick lookups, basic Q&A | | **Code Analysis** | `qwen/qwen3-coder` | $0.0000002 | Technical documentation, code review | | **Vision Tasks** | `z-ai/glm-4.5v` | $0.0000005 | Chart analysis, image interpretation | | **Complex Reasoning** | `x-ai/grok-4` | $0.000003 | Multi-step analysis, strategic planning | | **Code Editing** | `morph/morph-v3-large` | $0.0000009 | Fast code edits at 4500+ tokens/sec | ### **Caching Strategy** - **Result Caching**: 2-hour TTL with 85% similarity threshold - **Model Caching**: 1-hour TTL for repeated model responses - **Knowledge Base**: Automatic indexing of all research outputs - **Cost Savings**: 60-80% reduction through intelligent caching --- ## 🔗 **Tool Chaining Best Practices** ### **Health & Resilience Patterns** ```xml <best_practices> <health_monitoring> <step tool="get_server_status" frequency="startup" /> <step tool="db_health" frequency="daily" /> <step tool="list_models" frequency="weekly" refresh="true" /> </health_monitoring> <error_recovery> <step tool="get_job_status" condition="async_research" /> <step tool="cancel_job" condition="timeout_exceeded" /> <step tool="backup_db" condition="before_major_operations" /> </error_recovery> <quality_assurance> <step tool="get_past_research" purpose="context_validation" /> <step tool="rate_research_report" purpose="feedback_loop" /> <step tool="export_reports" purpose="audit_trail" /> </quality_assurance> </best_practices> ``` ### **Iterative Refinement Pattern** ```mcp # Initial broad research submit_research { "query": "AI alignment research", "costPreference": "low" } # Monitor progress get_job_status { "job_id": "[job_id]" } # Targeted follow-up based on initial findings research_follow_up { "originalQuery": "AI alignment research", "followUpQuestion": "What are the most promising current approaches to the alignment problem?" } # Cross-validate with existing knowledge search { "q": "AI alignment approaches", "scope": "reports", "rerank": true } # Quality feedback rate_research_report { "reportId": "[report_id]", "rating": 5, "comment": "Excellent comprehensive analysis" } ``` --- ## 🎯 **Manual Workflow Templates** For platforms without MCP prompt support, use these structured templates: ### **Template 1: Competitive Intelligence** ``` 1. Initial Search: search_web { "query": "[company/product] competitive analysis 2025" } 2. Source Verification: fetch_url { "url": "[credible_source]" } 3. Comprehensive Research: conduct_research { "query": "[detailed_analysis_prompt]", "textDocuments": ["source_content"] } 4. Follow-up Analysis: research_follow_up { "originalQuery": "[original]", "followUpQuestion": "[specific_aspect]" } 5. Quality Control: rate_research_report { "reportId": "[id]", "rating": [1-5] } ``` ### **Template 2: Technical Deep-Dive** ``` 1. Model Selection: list_models (identify technical/coding capable models) 2. Initial Research: conduct_research { "costPreference": "high", "audienceLevel": "expert" } 3. Knowledge Integration: search { "scope": "reports", "rerank": true } 4. Iterative Refinement: research_follow_up (2-3 iterations) 5. Documentation: backup_db (preserve findings) ``` ### **Template 3: Cost-Effective Bulk Research** ``` 1. Health Check: db_health + get_server_status 2. Batch Processing: submit_research (multiple async jobs) 3. Progress Monitoring: get_job_status (polling strategy) 4. Result Compilation: get_report_content (batch retrieval) 5. Quality Assessment: rate_research_report (feedback loop) ``` --- ## 💡 **Advanced Strategies** ### **Semantic Caching Optimization** - Use `search` tool before `conduct_research` to check for similar past work - Set `similarity threshold = 0.85` to balance freshness vs cost savings - Implement automatic cache warming for frequently researched topics ### **Model Selection Intelligence** - **Simple queries** → `deepseek/deepseek-chat-v3.1` (ultra-low cost) - **Vision tasks** → `z-ai/glm-4.5v` (multimodal capability) - **Code analysis** → `qwen/qwen3-coder` (specialized for technical content) - **Complex reasoning** → `x-ai/grok-4` (advanced reasoning capabilities) - **Code editing** → `morph/morph-v3-large` (4500+ tokens/sec edits) ### **Quality Assurance Framework** - **Pre-research**: Check `get_past_research` for context - **During research**: Monitor via `get_job_status` for async jobs - **Post-research**: Use `rate_research_report` for continuous improvement - **Maintenance**: Regular `backup_db` and `reindex_vectors` operations --- ## 🔧 **Production Deployment Checklist** ### **Configuration Verification** - [ ] `OPENROUTER_API_KEY` configured - [ ] `SERVER_API_KEY` for HTTP transport - [ ] `INDEXER_ENABLED=true` for knowledge base - [ ] `MCP_ENABLE_PROMPTS=true` and `MCP_ENABLE_RESOURCES=true` - [ ] Cost optimization models configured in environment ### **Health Monitoring** - [ ] `db_health` confirms database operational - [ ] `get_server_status` shows all systems ready - [ ] `list_models` returns comprehensive catalog - [ ] Sample `conduct_research` executes successfully ### **Performance Optimization** - [ ] Caching strategies enabled and configured - [ ] Parallelism tuned for infrastructure (default: 4) - [ ] Model selection algorithms validated - [ ] Cost thresholds appropriate for use case --- *Use these patterns as building blocks for sophisticated research workflows. Each pattern is designed for production reliability with comprehensive error handling and quality controls.*