# RAG MCP Client Configuration
**Generated:** 2026-01-23
**Database:** Agent Kiwi (mrecfyfjpwzrzxoiooew.supabase.co)
---
## Configuration Required
Add this to your MCP client configuration (Cursor/Claude settings):
```json
{
"mcpServers": {
"kiwi": {
"command": "python",
"args": ["-m", "kiwi_mcp"],
"env": {
"EMBEDDING_URL": "https://api.openai.com/v1/embeddings",
"EMBEDDING_API_KEY": "${YOUR_OPENAI_API_KEY}",
"EMBEDDING_MODEL": "text-embedding-3-small",
"VECTOR_DB_URL": "postgresql://postgres.mrecfyfjpwzrzxoiooew.supabase.co:5432/postgres",
"SUPABASE_URL": "https://mrecfyfjpwzrzxoiooew.supabase.co",
"SUPABASE_KEY": "${YOUR_SUPABASE_ANON_KEY}"
}
}
}
}
```
---
## Required Values to Replace
1. **`${YOUR_OPENAI_API_KEY}`** - Your OpenAI API key (starts with `sk-`)
- Get from: https://platform.openai.com/api-keys
2. **`${YOUR_SUPABASE_ANON_KEY}`** - Your Supabase anon/public key
- Get from: Supabase Dashboard → Project Settings → API → anon/public key
---
## OpenRouter Configuration (Alternative)
Prefer using OpenRouter? Here's the complete configuration:
```json
{
"mcpServers": {
"kiwi": {
"command": "python",
"args": ["-m", "kiwi_mcp"],
"env": {
"EMBEDDING_URL": "https://openrouter.ai/api/v1/embeddings",
"EMBEDDING_API_KEY": "${YOUR_OPENROUTER_API_KEY}",
"EMBEDDING_MODEL": "text-embedding-3-small",
"EMBEDDING_AUTH_HEADER": "Authorization",
"EMBEDDING_AUTH_FORMAT": "Bearer {key}",
"VECTOR_DB_URL": "postgresql://postgres.mrecfyfjpwzrzxoiooew.supabase.co:5432/postgres",
"SUPABASE_URL": "https://mrecfyfjpwzrzxoiooew.supabase.co",
"SUPABASE_KEY": "${YOUR_SUPABASE_ANON_KEY}"
}
}
}
}
```
**Get OpenRouter API key:** https://openrouter.ai/keys
---
## Environment Variables Explained
### Embedding Configuration (Mandatory)
- **`EMBEDDING_URL`**: OpenAI embeddings endpoint
- **`EMBEDDING_API_KEY`**: Your OpenAI API key
- **`EMBEDDING_MODEL`**: Embedding model to use (text-embedding-3-small is recommended)
### Vector Database (Mandatory)
- **`VECTOR_DB_URL`**: PostgreSQL connection string for pgvector
- Format: `postgresql://postgres.<ref>.supabase.co:5432/postgres`
- Your database ref: `mrecfyfjpwzrzxoiooew`
### Supabase Registry (Optional but Recommended)
- **`SUPABASE_URL`**: Your Supabase project URL
- **`SUPABASE_KEY`**: Anon/public key for registry access
---
## Alternative Embedding Providers
### OpenRouter
```json
{
"EMBEDDING_URL": "https://openrouter.ai/api/v1/embeddings",
"EMBEDDING_API_KEY": "${YOUR_OPENROUTER_API_KEY}",
"EMBEDDING_MODEL": "text-embedding-3-small",
"EMBEDDING_AUTH_HEADER": "Authorization",
"EMBEDDING_AUTH_FORMAT": "Bearer {key}"
}
```
**Supported models:**
- `text-embedding-3-small` (OpenAI via OpenRouter)
- `text-embedding-3-large` (OpenAI via OpenRouter)
- `text-embedding-ada-002` (OpenAI via OpenRouter)
**Get API key:** https://openrouter.ai/keys
**Pricing:** Same as OpenAI but with unified billing across providers
### Ollama (Local)
```json
{
"EMBEDDING_URL": "http://localhost:11434/api/embeddings",
"EMBEDDING_API_KEY": "not-needed",
"EMBEDDING_MODEL": "nomic-embed-text",
"EMBEDDING_FORMAT": "ollama"
}
```
### Azure OpenAI
```json
{
"EMBEDDING_URL": "https://<your-resource>.openai.azure.com/openai/deployments/<deployment-name>/embeddings?api-version=2023-05-15",
"EMBEDDING_API_KEY": "${YOUR_AZURE_KEY}",
"EMBEDDING_MODEL": "text-embedding-ada-002",
"EMBEDDING_AUTH_HEADER": "api-key",
"EMBEDDING_AUTH_FORMAT": "{key}"
}
```
---
## Verification
After adding the configuration, verify it works:
```bash
# The MCP server should start without errors
python -m kiwi_mcp
# Expected: Server starts successfully
# If RAG config is missing, you'll see an error with instructions
```
---
## What Happens Next
Once configured, the Kiwi MCP will:
1. ✅ **Validate RAG config** at startup (fails fast if missing)
2. ✅ **Automatically embed** all validated content (directives, tools, knowledge)
3. ✅ **Enable semantic search** across all content types
4. ✅ **Use hybrid search** (vector + keyword) for better results
---
## Database Tables Created
Your database now has:
| Table | Purpose | Rows |
|-------|---------|------|
| `item_embeddings` | Vector embeddings for all content | 0 (ready) |
| `knowledge` | Knowledge metadata | 159 |
| `knowledge_versions` | Knowledge content versions | 159 |
| `directives` | Directive metadata | 141 |
| `directive_versions` | Directive content versions | 351 |
| `tools` | Tool metadata | 15 |
| `tool_versions` | Tool content versions | 15 |
---
## Cost Estimation
**OpenAI text-embedding-3-small pricing:**
- ~$0.02 per 1M tokens
- Average knowledge entry: ~500 tokens
- Your 159 entries: ~79,500 tokens
- **Estimated cost**: ~$0.002 (less than a penny!)
**Ongoing costs:**
- Each new/updated item: ~$0.00001
- Queries are free (just retrieval)
---
## Troubleshooting
### "RAG configuration missing" error
If you see this error, the environment variables aren't being passed to the MCP server.
**Solution**: Make sure your MCP client configuration is in the correct location:
- **Cursor**: Settings → MCP → Add server configuration
- **Claude Desktop**: `~/Library/Application Support/Claude/claude_desktop_config.json` (Mac) or `%APPDATA%\Claude\claude_desktop_config.json` (Windows)
### Connection errors
If you see database connection errors:
1. Check your Supabase database is **active** (not paused)
2. Verify the connection string is correct
3. Test the connection:
```bash
psql postgresql://postgres.mrecfyfjpwzrzxoiooew.supabase.co:5432/postgres
```
### Embedding errors
If embeddings fail:
1. Verify your OpenAI API key is valid
2. Check you have credits available
3. Try the embedding endpoint manually:
```bash
curl https://api.openai.com/v1/embeddings \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"input": "test", "model": "text-embedding-3-small"}'
```
---
## Related Documentation
- [RAG_INTEGRATION_COMPLETE.md](./RAG_INTEGRATION_COMPLETE.md) - Complete RAG implementation
- [RAG_IMPLEMENTATION_SUMMARY.md](./RAG_IMPLEMENTATION_SUMMARY.md) - Technical summary
- [DATABASE_SCHEMA_ALIGNMENT.md](./DATABASE_SCHEMA_ALIGNMENT.md) - Database structure
