The Model Context Protocol (MCP) lets AI assistants connect directly to your knowledge graph. Search, explore relationships, and write insights back - all through natural conversation.
MCP is an open standard that lets AI assistants connect to external data sources and tools. Graphory provides an MCP server that gives your AI read and write access to your knowledge graph.
Instead of copy-pasting data into chat, your AI can directly search your graph, look up specific items, explore connections, and save findings back for future reference.
See the Getting Started guide for platform-specific configuration instructions covering Claude Desktop, Claude Code, ChatGPT, Gemini CLI, Cursor, Windsurf, VS Code, Codex CLI, and direct API access.
Once connected, your AI has access to these tools:
Search your knowledge graph for items matching a query. Supports filtering by type and business entity.
| Parameter | Description |
|---|---|
query | Search terms |
node_type | Optional type filter (e.g. "person", "company", "email") |
entity | Optional business entity filter |
limit | Max results (default: 20) |
Get full details for a specific item including all its properties and direct relationships.
| Parameter | Description |
|---|---|
node_id | The item's unique identifier |
Explore the relationship network starting from a specific item. Reveals connections up to N hops away.
| Parameter | Description |
|---|---|
start_id | Item ID to start from |
depth | Hops to traverse (default: 2, max: 5) |
edge_types | Optional relationship type filter |
Get a chronological feed of recent activity across your data sources.
| Parameter | Description |
|---|---|
entity | Optional business entity filter |
days | Lookback window (default: 30) |
limit | Max items (default: 50) |
Get a summary overview of your graph - total counts, breakdowns by type and source.
No parameters required.
Save findings or new information back to the graph. Confidence gating applies (auto-merge at 0.9+, review queue at 0.7-0.89, rejected below 0.7).
| Parameter | Description |
|---|---|
action | add, connect, or correct |
confidence | Confidence score (0 to 1) |
data | JSON payload for the write operation |
Push raw data into your graph from within a conversation. Useful for capturing insights on the fly.
| Parameter | Description |
|---|---|
payload | JSON string with entity, source, title, and body fields |
Who are the key contacts at Acme Corp?
search_graph to find Acme Corp, then traverse to discover connected people. It combines results from emails, calendar events, and other sources into a unified answer.
What happened in the last week across all my business entities?
timeline with a 7-day window and summarizes activity across all connected data sources.
Based on those emails, it looks like Jane Smith just moved from Acme to Beta Inc. Update the graph.
write_to_graph to update Jane's company association with high confidence, since you confirmed the change.
What does my graph look like? Give me the overview.
graph_stats to get total node/edge counts and breakdowns by type and source, then presents a summary.
gs_ak_node_type, entity) to narrow search scopedepth on traverse operations if results are too broad