Documentation
Everything you need to connect your AI agent to the Kohnex Brain knowledge graph.
Quickstart
Add Kohnex Brain as a remote MCP server in your OpenCode config. You'll need an API key from the dashboard.
{
"mcp": {
"kohnex": {
"type": "remote",
"url": "https://brain.kohnex.com/mcp",
"headers": {
"Authorization": "Bearer khnx_<your_key>"
}
}
}
}
Then restart OpenCode. The kohnex tools will be available in any agent.
Setup
1. Get an API Key
Sign up at kohnex.com/pricing — the free Starter plan gives you 100 queries to try the brain.
2. Add to OpenCode
Paste the config snippet above into your ~/.config/opencode/opencode.json file.
3. Start Querying
Use any Kohnex tool in your prompts: use kohnex to find nature metaphors for consensus algorithms
MCP Tools
Five tools, each exposing a different facet of the knowledge graph. kohnex_path and kohnex_domain require a Pro subscription.
kohnex_think
— Full brain activation
Query with optional mode. Returns top activated nodes, nature metaphor, ancient wisdom, and eureka connections.
kohnex_path
— Shortest path search
Find the shortest path between any two concepts in the graph.
kohnex_domain
— Domain explorer
List all nodes in a domain: cyber, fault, energy, performance, compression, collective, signal, memory, distributed.
kohnex_mode
— Mode description
Describe a thinking mode's strategy, depth, and best use cases.
kohnex_stats
— Brain statistics
Get overview: node counts by type, edge distribution, inhibitory ratio, domain coverage.
Thinking Modes
Each mode changes the brain's activation strategy — depth, threshold, and inhibition behavior.
| Mode | Strategy | Best For |
|---|---|---|
| Focus | Deep, high-threshold | Precision answers |
| Explore | Broad, low-threshold | Discovery & learning |
| Out-of-Box | Lateral, weak-tone | Creative problem solving |
| Dream | Randomized, low-inhibition | Serendipitous connections |
| Intuitive | Fast pattern match | Quick analogies |
| Beamform | Multi-seed intersection | Complex multi-concept queries |
| Swarm | Collective consensus | Multi-agent reasoning |
| Predator | Adversarial search | Security threat analysis |
| Regeneration | Self-healing circuits | Fault tolerance design |
| Hibernate | Low-energy | Energy optimization |
| Immune | Threat pattern matching | Security defense |
| Photosynthesis | Energy harvesting | Sustainable compute |
| Quantum | Superposition reasoning | Uncertainty & probability |
Example Queries
Security architecture
Activates: Biofilm (5-layer defense), Stentor (escalating response), CRISPR (adaptive immune memory), Defense-in-Depth
GPU optimization
Activates: Retina Preprocessing (100:1 compression), Neural Sparse Coding (1-4% active), DNA multi-level packaging, HBM
Distributed consensus
Activates: Nash Equilibrium, Bee Hive Democracy, Quorum Sensing, Prisoner's Dilemma, Firefly Synchronization
FAQ
What MCP clients work?
Any MCP-compatible client: OpenCode, Claude Desktop, Cursor, VS Code with MCP extension, or custom clients using the MCP SDK.
Is the knowledge graph updated?
Yes — the graph is actively maintained. New nodes and edges are added regularly. The stats endpoint always shows the current counts.
Can I add my own knowledge?
Enterprise customers can extend the graph with custom nodes and edges. Contact us for details.