Integration
Custom Agent Integration
Add ekkOS memory to your own AI agents and applications.
Integration Options
There are two ways to integrate ekkOS with custom agents:
MCP Protocol
For agents that support the Model Context Protocol
REST API
Direct HTTP calls for any application
REST API Integration
Search Memory
Find relevant patterns and context:
curl -X POST https://mcp.ekkos.dev/api/v1/memory/search \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"query": "authentication patterns",
"limit": 5
}'Save Pattern
Store a new pattern:
curl -X POST https://mcp.ekkos.dev/api/v1/patterns \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"title": "Rate limiting with Redis",
"problem": "Need to limit API requests",
"solution": "Use Redis sorted sets with sliding window",
"tags": ["api", "redis", "performance"]
}'Write Working Memory
Capture conversation messages:
curl -X POST https://mcp.ekkos.dev/api/v1/working \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"role": "user",
"content": "Help me implement user authentication",
"session_id": "session_123",
"source": "my-custom-agent"
}'TypeScript SDK
Installation
npm install @ekkos/sdk
Usage Example
import { EkkosClient } from '@ekkos/sdk';
const ekkos = new EkkosClient({
apiKey: process.env.EKKOS_API_KEY
});
// Search memory
const results = await ekkos.search({
query: 'authentication best practices',
limit: 5
});
// Inject context into your AI prompt
const context = results.map(r => r.content).join('\n');
const prompt = `
Given this context from memory:
${context}
User question: ${userQuestion}
`;
// Save a new pattern
await ekkos.forgeInsight({
title: 'JWT refresh token rotation',
problem: 'Tokens expire during long sessions',
solution: 'Implement automatic token refresh with rotation',
tags: ['auth', 'security', 'jwt']
});
// Track outcome
await ekkos.recordOutcome({
applicationId: results.retrieval_id,
success: true
});Recommended Integration Pattern
- 1Before AI call — Search memory for relevant context
- 2Inject context — Add retrieved patterns to system prompt
- 3Make AI call — Let the AI use the enriched context
- 4Capture conversation — Write messages to working memory
- 5Track outcome — Report if the patterns helped