AI Setup with Anthropic¶

Section: AI Assistant | Article 29
Audience: System Administrators
Last Updated: 2026-04-07

Overview¶

This article walks through configuring Anthropic (Claude) as the AI completion provider for RP-PAM. Anthropic provides high-quality language model completions through the Claude family of models.

Important: Anthropic does not currently offer a standalone embedding API. If you use Anthropic for completions, you must pair it with a second provider (Ollama or OpenAI) for embedding support. Without embeddings, risk scoring and anomaly detection will not function.

Prerequisites¶

Requirement	Detail
RP-PAM licence	Enterprise or MSP tier
Anthropic account	An account at console.anthropic.com
API key	A valid Anthropic API key
Network access	The RP-PAM server must be able to reach `api.anthropic.com` on port 443
Embedding provider	Ollama or OpenAI configured for embeddings (see note above)

Step 1: Get an Anthropic API Key¶

Sign in to console.anthropic.com.
Navigate to API Keys in the settings.
Click Create Key.
Name it: RP-PAM
Copy the key immediately.

The key looks like: sk-ant-api03-...

Step 2: Encrypt the API Key¶

Store the API key in RP-PAM's encrypted vault.

Windows PowerShell:

& "C:\Program Files\Ravenphyre\RP-PAM\tools\rppam-migrate.exe" vault-store `
  --key "ai-anthropic-key" `
  --value "sk-ant-api03-YOUR_ANTHROPIC_API_KEY"

Linux:

sudo /opt/rppam/tools/rppam-migrate vault-store \
  --key "ai-anthropic-key" \
  --value "sk-ant-api03-YOUR_ANTHROPIC_API_KEY"

Step 3: Configure rppam.config¶

Option A: Anthropic Only (Completions Without Embeddings)¶

If you do not need risk scoring or anomaly detection:

{
  "ai": {
    "enabled": true,
    "provider": "anthropic",
    "apiKeyVaultKey": "ai-anthropic-key",
    "completionModel": "claude-sonnet-4-20250514",
    "maxTokens": 4096,
    "temperature": 0.3,
    "riskScoring": {
      "enabled": false
    },
    "anomalyDetection": {
      "enabled": false
    }
  }
}

This configuration enables natural-language queries and access request parsing, but disables features that require embeddings.

Option B: Anthropic + Ollama (Full Features, Recommended)¶

Pair Anthropic completions with Ollama embeddings for full functionality with maximum privacy on the embedding side:

{
  "ai": {
    "enabled": true,
    "provider": "anthropic",
    "apiKeyVaultKey": "ai-anthropic-key",
    "completionModel": "claude-sonnet-4-20250514",
    "maxTokens": 4096,
    "temperature": 0.3,
    "embedding": {
      "provider": "ollama",
      "ollamaBaseUrl": "http://localhost:11434",
      "embeddingModel": "nomic-embed-text",
      "embeddingDimension": 768
    },
    "riskScoring": {
      "enabled": true,
      "threshold": 60
    },
    "anomalyDetection": {
      "enabled": true,
      "lookbackDays": 90
    }
  }
}

Option C: Anthropic + OpenAI (Full Features, Cloud)¶

Pair Anthropic completions with OpenAI embeddings:

{
  "ai": {
    "enabled": true,
    "provider": "anthropic",
    "apiKeyVaultKey": "ai-anthropic-key",
    "completionModel": "claude-sonnet-4-20250514",
    "maxTokens": 4096,
    "temperature": 0.3,
    "embedding": {
      "provider": "openai",
      "apiKeyVaultKey": "ai-openai-key",
      "embeddingModel": "text-embedding-3-small",
      "embeddingDimension": 1536
    },
    "riskScoring": {
      "enabled": true,
      "threshold": 60
    },
    "anomalyDetection": {
      "enabled": true,
      "lookbackDays": 90
    }
  }
}

Note: Option C requires storing a second API key for OpenAI (see Article 28 Steps 1-2).

Configuration Fields¶

Field	Description	Default
`provider`	Set to `"anthropic"`	(required)
`apiKeyVaultKey`	Vault key for the Anthropic API key	(required)
`completionModel`	Claude model for completions	`claude-sonnet-4-20250514`
`maxTokens`	Maximum tokens in a completion response	`4096`
`temperature`	Response creativity (0.0-1.0)	`0.3`
`embedding.provider`	Separate provider for embeddings: `"ollama"` or `"openai"`	(optional)
`embedding.embeddingModel`	Embedding model name	(depends on provider)
`embedding.embeddingDimension`	Vector dimension for the embedding model	(depends on model)

Recommended Anthropic Models¶

Model	Use Case	Notes
`claude-sonnet-4-20250514`	General purpose (recommended)	Best balance of quality and cost
`claude-opus-4-20250514`	Complex reasoning	Higher quality, higher cost
`claude-haiku-3-20250307`	Fast, cost-effective	Good for high-volume, simpler queries

Step 4: Restart RP-PAM¶

Windows PowerShell:

Restart-Service RpPam

Linux:

sudo systemctl restart rppam

Step 5: Verify AI Is Working¶

Check Module Health¶

Linux:

curl -s http://localhost:7101/api/v1/modules \
  -H "Authorization: Bearer $ADMIN_JWT" | jq '.items[] | select(.moduleName == "ai")'

PowerShell:

$modules = Invoke-RestMethod -Uri "http://localhost:7101/api/v1/modules" `
  -Headers @{ Authorization = "Bearer $adminJwt" }
$modules.items | Where-Object { $_.moduleName -eq "ai" } | ConvertTo-Json

Expected:

{
  "moduleName": "ai",
  "status": "healthy",
  "provider": "anthropic",
  "completionModel": "claude-sonnet-4-20250514",
  "embeddingProvider": "ollama",
  "embeddingModel": "nomic-embed-text"
}

Test a Query¶

Linux:

curl -s -X POST http://localhost:7101/api/v1/ai/query \
  -H "Authorization: Bearer $ADMIN_JWT" \
  -H "Content-Type: application/json" \
  -d '{ "query": "Summarise access activity for today." }' | jq .

PowerShell:

$response = Invoke-RestMethod -Uri "http://localhost:7101/api/v1/ai/query" `
  -Method Post `
  -Headers @{ Authorization = "Bearer $adminJwt" } `
  -ContentType "application/json" `
  -Body '{ "query": "Summarise access activity for today." }'
$response | ConvertTo-Json

Troubleshooting¶

Problem	Cause	Solution
`"status": "unhealthy"` for AI module	API key invalid or expired	Regenerate the key at console.anthropic.com; re-store with `vault-store`
`401` or `authentication_error`	Wrong API key	Verify the vault key name matches `apiKeyVaultKey`
`429 rate_limit_error`	Rate limit exceeded	Reduce request frequency or contact Anthropic for higher limits
Risk scoring shows "not available"	No embedding provider configured	Add an `embedding` section with Ollama or OpenAI
Slow responses	Large model or network latency	Try `claude-haiku-3-20250307` for faster responses
`overloaded_error`	Anthropic API temporarily overloaded	Retry after a few seconds; this is transient

Next Steps¶

AI Setup with Ollama -- Set up Ollama for self-hosted embeddings (pairs well with Anthropic)
AI Setup with OpenAI -- Alternative provider with built-in embeddings
AI Assistant Overview -- What the AI module can do