Agent Deploy Dashboard MCP

What aparajithn/agent-deploy-dashboard-mcp MCP server does

The aparajithn/agent-deploy-dashboard-mcp MCP server presents deployment operations for Vercel, Render, Railway, and Fly.io through a shared interface. An agent can list services across platforms, inspect deployment state, retrieve logs, manage environment variables, trigger redeployments, and check the health of a URL. The server also exposes the same core functionality through a REST API.

The tool set includes list_all_services, get_deploy_status, tail_logs, get_env_vars, set_env_var, trigger_redeploy, get_build_logs, check_health, and rollback_deploy. Inputs generally identify a platform and service, while log retrieval accepts a line count and health checks accept a URL. Deployment log retrieval uses a deployment ID, and rollback accepts a target version or deployment ID.

How it works

The hosted MCP endpoint is available at https://agent-deploy-dashboard-mcp.onrender.com/mcp. An MCP client can register that URL as a remote server. The REST API uses the corresponding https://agent-deploy-dashboard-mcp.onrender.com base URL, with endpoints under /api/v1/ for individual operations.

Platform access is configured with provider tokens. The server uses separate credentials for Vercel, Render, Railway, and Fly.io, allowing requests to be routed to the selected provider. Results use a common response structure containing fields such as success status, platform, service ID, deployment ID, logs, or health information, depending on the operation.

The service applies a free quota of 50 requests per IP address per day. After that quota, the documented paid path uses HTTP 402 payments at $0.01 per request and requires payment proof in an X-Payment header. The README identifies this as x402 payment middleware.

Setup and configuration

For a remote MCP client, add the hosted /mcp URL to the client's MCP server configuration. The README specifically shows this pattern for Cline settings. Local development uses Python: clone the repository, create a virtual environment, activate it, and install the project with pip install -e ".[dev]".

Configure provider credentials through environment variables:

• VERCEL_TOKEN for Vercel access
• RENDER_API_KEY for Render access
• RAILWAY_TOKEN for Railway access
• FLY_API_TOKEN for Fly.io access

The README also documents X402_WALLET_ADDRESS for payment configuration, marked as optional. The hosted endpoint and local server therefore have different setup paths: remote clients can use the published URL, while a local deployment needs the project and provider credentials.

Tools and capabilities

The aparajithn/agent-deploy-dashboard-mcp MCP server supports these operations:

• List services from all supported platforms.
• Check deployment status for a specified service.
• Tail recent service logs with a configurable number of lines.
• Read or set environment variables.
• Trigger a redeployment.
• Retrieve build logs for a deployment.
• Check an HTTP health endpoint and report status code, health result, response time, and headers.
• Request rollback to a previous deployment version.

Vercel and Render are documented as having full support for service listing, deployment status, logs, environment variables, redeployment, and build logs, except that Vercel service logs are limited to build logs. Railway and Fly.io currently support basic service listing; their other listed operations are marked planned or stub implementations.

Limitations and notes

Coverage is not uniform across providers. In particular, Railway and Fly.io do not yet have the full deployment-management surface described for Vercel and Render. Rollback is exposed as a tool, but the example response indicates that rollback may not be implemented for a given platform.

Environment-variable operations can expose or change deployment configuration, so agents should receive only the provider credentials and service identifiers appropriate to their task. The free tier is quota-based rather than unlimited, and requests beyond the quota require the documented payment flow. The material does not specify a local command for launching the MCP server itself, so the published remote endpoint is the clearest installation option for compatible clients.