BUILT FOR AI AGENTS

Give every Agent an execution decision layer.

ModelPilot lets an Agent ask which execution path is allowed, run the work, verify the outcome, and retain auditable evidence—with one tenant API key.

Create an API key Download VS Code config

Task → Policy Filter
→ Required Intelligence
→ Execution Target
→ Execute → Verify → Learn

Not Ollama. Not OpenRouter.
No model hosting or model-access resaleControl plane

THREE WAYS TO CONNECT

Bind once. Keep your providers.

Your Agent and your provider credentials stay under your control. ModelPilot supplies policy-aware decisions, verified execution, evidence, and task economics.

Remote MCP

Best for MCP-capable agents. Discover ModelPilot tools over one Streamable HTTP endpoint.

MCP details →

Local CLI

Install the CLI locally to configure credentials, run diagnostics, preview decisions, and execute verified tasks.

npm run cli:install

Native Task API

Use OpenAPI when your application needs direct control over policy, economics, and verification contracts.

OpenAPI document →

VS CODE MCP

Use a secure prompt for the API key.

Save this as .vscode/mcp.json. VS Code asks for the key and treats it as a password instead of committing it to the repository.

{
  "inputs": [
    {
      "type": "promptString",
      "id": "modelpilot-api-key",
      "description": "ModelPilot API key",
      "password": true
    }
  ],
  "servers": {
    "modelpilot": {
      "type": "http",
      "url": "https://modelpilot.andycywu.cc/mcp",
      "headers": {
        "Authorization": "Bearer ${input:modelpilot-api-key}"
      }
    }
  }
}

AGENT OPERATING CONTRACT

Make verified work the default.

Decide

Call modelpilot_decide_execution before choosing an execution target.

Execute and verify

Use modelpilot_execute_verified_task for complete work and do not report success unless the outcome is verified.

Attribute honestly

Never infer realized value. Confirm it only from customer input or a verifiable system of record.