@runtimeai/mcp-server-aep

MCP Server for the RuntimeAI Agentic Enablement Platform (AEP) — 49 tools across 8 categories covering agent identity (KYA), PII tokenization, memory vault, cost control, audit, and fraud detection.

Installation

npm install -g @runtimeai/mcp-server-aep
# or run directly
npx @runtimeai/mcp-server-aep

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | AEP_API_URL | Yes | AEP service base URL, e.g. https://aep.rt19.runtimeai.io | | AEP_API_KEY | Yes* | API key for authenticated tools | | AEP_TENANT_ID | Yes* | Tenant ID for multi-tenant requests |

*Required for all tools except trial_signup and trial_verify.

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "aep": {
      "command": "npx",
      "args": ["-y", "@runtimeai/mcp-server-aep"],
      "env": {
        "AEP_API_URL": "https://aep.rt19.runtimeai.io",
        "AEP_API_KEY": "your-api-key-here",
        "AEP_TENANT_ID": "your-tenant-id-here"
      }
    }
  }
}

Tools Reference (49 tools)

KYA — Know Your Agent (8 tools)

| Tool | Description | |------|-------------| | kya_score_agent | Score an agent's trustworthiness and compliance posture | | kya_register_agent | Register a new agent in the KYA registry | | kya_get_agent | Retrieve full details for a registered agent | | kya_list_agents | List all registered agents in the tenant | | kya_block_agent | Block an agent from operating | | kya_update_agent | Update agent metadata (name, owner, tags) | | kya_get_score_history | Retrieve historical trust scores for an agent | | kya_bulk_score | Score multiple agents in a single request |

PII Shield (8 tools)

| Tool | Description | |------|-------------| | pii_tokenize | Tokenize a single PII value | | pii_detokenize | Resolve a PII token back to its original value | | pii_tokenize_batch | Tokenize multiple PII values in one call | | pii_detokenize_batch | Resolve multiple PII tokens in one call | | pii_scan_text | Scan free-form text for PII entities | | pii_redact_text | Redact all detected PII from text | | pii_get_token_info | Get metadata for a PII token | | pii_get_vault_stats | Get PII vault statistics |

Memory Vault (8 tools)

| Tool | Description | |------|-------------| | memory_store | Store a value in the agent memory vault | | memory_retrieve | Retrieve a value from the memory vault by key | | memory_list_keys | List all memory keys in the vault | | memory_delete | Delete a key from the memory vault | | memory_search | Semantic search across memory vault contents | | memory_append | Append a value to an existing memory entry | | memory_get_stats | Get memory vault statistics | | memory_export | Export all memory vault entries |

Cost Control (6 tools)

| Tool | Description | |------|-------------| | cost_set_budget | Create or update an agent budget | | cost_get_budget | Retrieve budget details and current spend | | cost_list_budgets | List all configured budgets | | cost_get_usage | Query token usage and spend metrics | | cost_get_alerts | List active cost alerts | | cost_record_usage | Record a usage event for cost attribution |

Audit Black Box (6 tools)

| Tool | Description | |------|-------------| | audit_query | Query the immutable audit log | | audit_get_event | Retrieve a single audit event with proof | | audit_verify_chain | Verify audit chain integrity for a time window | | audit_get_stats | Get audit statistics and volume metrics | | audit_export | Export audit events to JSON or CSV | | audit_log_event | Manually log a custom audit event |

Fraud Shield (5 tools)

| Tool | Description | |------|-------------| | fraud_check_email | Check an email for fraud signals | | fraud_check_domain | Analyze a domain for fraud indicators | | fraud_check_agent | Run fraud detection on an agent | | fraud_check_batch | Check multiple items for fraud at once | | fraud_get_report | Get a fraud detection summary report |

Agent Orchestration (5 tools)

| Tool | Description | |------|-------------| | orchestration_run_job | Submit an orchestration job | | orchestration_get_job | Get job status and results | | orchestration_list_jobs | List orchestration jobs | | orchestration_cancel_job | Cancel a running or queued job | | orchestration_run_full_audit | Run full KYA + PII + audit chain verification |

Trial (3 tools — no authentication required)

| Tool | Description | |------|-------------| | trial_signup | Sign up for a free AEP trial | | trial_verify | Activate trial using magic link token | | trial_status | Check trial status and usage |

Example Usage

Register and score an agent

Use kya_register_agent to register a new agent named "SalesBot" owned by "[email protected]" with type "llm"

Then use kya_score_agent with the returned agent_id to check its trust score

Protect PII before sending to an LLM

Use pii_redact_text with the text "Customer John Smith ([email protected], SSN 123-45-6789) called today" to get a safe version to pass to Claude

Set a cost budget for an agent

Use cost_set_budget with agent_id "agent_abc123", daily_limit 5.00, and monthly_limit 100.00 to cap spending for a specific agent

Query recent audit events

Use audit_query with start_time "2026-05-01T00:00:00Z" and end_time "2026-05-20T23:59:59Z" to see all events this month

Local Development

cd mcp-servers/mcp-server-aep
npm install
npm run build
AEP_API_URL=http://localhost:8301 AEP_API_KEY=dev-key AEP_TENANT_ID=dev-tenant npm start

MCP Inspector

npm run inspector

Architecture

All 49 tools route through a single api() helper that:

1. Prepends AEP_API_URL to the path
2. Injects X-AEP-API-Key and X-AEP-Tenant-ID headers
3. Returns normalized JSON (success body or { error, status })

Trial tools use a separate trialApi() helper that sends no auth headers.

License

MIT — RuntimeAI, Inc.