MCP Usage Examples
Real-world workflows using ExoGraph MCP.
Research Assistant
Scenario: You're researching transformer architectures and want to build a knowledge base.
You: "Search my ExoGraph for transformer architectures"
Claude: [Shows 5 relevant entities from your graphs]
You: "Tell me more about the top result"
Claude: [Uses get_world_graph to show relationships]
You: "We're missing recent developments. Start a research session"
Claude: [Creates research session, provides session ID]
You: "Check the status"
Claude: [Shows research is running, X sources found so far]Knowledge Discovery
Scenario: Exploring connections in your knowledge graph.
You: "List my ExoGraph graphs"
Claude: [Shows 3 graphs you have]
You: "Search the ML papers graph for attention mechanisms"
Claude: [Searches specific graph, returns entities]
You: "How does 'attention' relate to 'transformers'?"
Claude: [Shows relationship graph with evidence]Document Q&A
Scenario: You've uploaded papers and want to query them.
You: "I just uploaded a quantum computing paper to ExoGraph.
Search for quantum entanglement"
Claude: [Searches and finds entities from your new document]
You: "Show me how this connects to my existing knowledge"
Claude: [Shows relationships in the world graph]
You: "Summarize the key insights"
Claude: [Uses context from search results to summarize]Automated Research
Scenario: Let AI do the research for you.
You: "Start ExoGraph research on 'quantum error correction methods'
with a budget of 100 credits"
Claude: [Creates research session with budget limit]
You: "What's the status?"
Claude: [Shows progress - found 5 sources, analyzing...]
... wait 2 minutes ...
You: "Check again"
Claude: [Shows completed research with summary and sources]
You: "Search my graphs for the concepts mentioned in the research"
Claude: [Searches and shows existing related knowledge]Multi-Graph Analysis
Scenario: Comparing knowledge across different graphs.
You: "List my ExoGraph graphs"
Claude: [Shows: Work Papers, Personal Research, Course Notes]
You: "Search 'neural networks' in Work Papers"
Claude: [Shows results from work graph]
You: "Now search the same in Personal Research"
Claude: [Shows results from personal graph]
You: "Are there any connections between these results?"
Claude: [Uses get_world_graph to find cross-graph connections]Continuous Learning
Scenario: Build knowledge over time.
Day 1:
You: "Start ExoGraph research on 'GPT architecture'"
Claude: [Creates research session, completes]
Day 2:
You: "Search my graphs for GPT"
Claude: [Finds yesterday's research results]
You: "Now research 'GPT training techniques'"
Claude: [New research builds on existing knowledge]
Day 3:
You: "How does 'GPT training' relate to 'GPT architecture'?"
Claude: [Shows connections from your accumulated research]Literature Review
Scenario: Systematic review of academic papers.
You: "I'm reviewing papers on attention mechanisms.
List my ExoGraph graphs"
Claude: [Shows available graphs]
You: "Search for 'self-attention' across all graphs"
Claude: [Searches and returns entities with papers]
You: "Show me the relationships between the top 3 results"
Claude: [Uses get_world_graph for each, shows connections]
You: "Start research on topics we're missing coverage for"
Claude: [Identifies gaps, starts research session]Concept Exploration
Scenario: Understanding a new technical concept.
You: "Search my ExoGraph for 'knowledge distillation'"
Claude: [Returns entities, some results found]
You: "Not enough information. Research this topic"
Claude: [Starts research session]
You: [Wait a minute] "Check the research status"
Claude: [Shows progress, found 10 sources]
You: [After completion] "How does knowledge distillation
relate to other ML techniques in my graph?"
Claude: [Shows new connections from research results]Collaborative Research
Scenario: Team using shared ExoGraph account.
Morning:
You: "List ExoGraph graphs"
Claude: [Shows team's shared graphs]
You: "Search 'quantum algorithms' in the Team Research graph"
Claude: [Shows what teammates have added]
Afternoon:
You: "Start research on 'variational quantum eigensolver'"
Claude: [Adds to shared knowledge]
Next Day:
Teammate asks: "Search for VQE"
Claude: [Finds your research from yesterday]Best Practices
1. Be Specific
Good: "Search my ExoGraph ML papers for attention mechanisms"
Better: "Search my ExoGraph for self-attention in transformers"
2. Use Context
Claude remembers the conversation:
You: "Search my ExoGraph for quantum computing"
Claude: [Shows results with entity IDs]
You: "Show relationships for the top result"
Claude: [Uses the entity ID from previous results]3. Chain Actions
"Search for X, then show me how it relates to Y,
and if we don't have enough information, start research"4. Specify Graphs
When you have multiple graphs:
"Search my 'Work Papers' graph in ExoGraph for X"5. Monitor Research Progress
You: "Start research on X"
[Wait 30 seconds]
You: "Check the status"
[Repeat until complete]6. Use Budget Limits
"Start research on X with a budget of 50 credits"Prevents unexpected charges for expensive research.
Tips & Tricks
Get Session IDs
You: "Start research on X"
Claude: [Provides session ID in response]
You: "What was that session ID?"
Claude: [Repeats it from conversation history]Batch Operations
You: "Search for 'quantum', 'neural networks', and 'blockchain'
in my ExoGraph"
Claude: [Runs multiple searches, summarizes all results]Compare Results
You: "Search for 'transformers' in my 2023 papers and 2024 papers"
Claude: [Searches both, highlights differences in understanding]Extract Insights
You: "Search my ExoGraph for recent AI safety research"
Claude: [Shows results]
You: "What are the common themes?"
Claude: [Analyzes entity relationships, identifies patterns]Advanced Workflows
Weekly Knowledge Digest
Monday:
"List all ExoGraph graphs created this week"
"Search each for my key research topics"
"Summarize new insights"Gap Analysis
"Search my ExoGraph for [Topic A]"
"Search for [Topic B]"
"Are there any connections? If not, research the intersection"Progressive Research
1. "Search for broad topic"
2. "Find subtopics with few connections"
3. "Research those subtopics"
4. "Repeat until comprehensive coverage"Integration with Other Tools
With Web App
- Upload documents via web app
- Query via MCP in Claude Desktop
- View visualizations in web app
- Refine queries in Claude Desktop
With REST API
- Use MCP for exploration
- Note useful patterns
- Implement in production via REST API
- Use MCP for debugging
Real User Stories
Academic Researcher
"I use MCP to explore my literature database. Instead of clicking through papers, I just ask Claude 'how does paper X relate to Y' and get instant answers."
ML Engineer
"When learning new concepts, I upload papers and ask Claude to find connections. The research feature fills gaps in my knowledge automatically."
Knowledge Manager
"Our team collaborates through ExoGraph. MCP makes it easy to query our shared knowledge base without learning the API."
Common Patterns
Pattern 1: Search → Explore → Research
Search for existing knowledge
→ Explore relationships
→ Research gapsPattern 2: Upload → Query → Synthesize
Upload documents via web app
→ Query via MCP
→ Let Claude synthesize insightsPattern 3: Monitor → Act → Verify
List graphs to see what's available
→ Take action (search/research)
→ Verify results with follow-up queriesNext Steps
- Tool Reference - Detailed tool docs
- Troubleshooting - Having issues?
- REST API - For programmatic access
- Web App - Upload & visualize