Render MCP
Official Render server for managing services, deploys, logs, metrics and Postgres from an AI client
BestMCPTools
Render MCP
bestmcptools.org
Last verified: November 10, 2026
Maintenance status
ActiveNot archived. 2026 signals: Render changelog entries dated July 17, 2026 (trigger deploys with the Render MCP server) and July 22, 2026 (OAuth for Claude Code, Codex and Cursor). Latest GitHub release v0.3.0 shows 14 January with no year on the page. Commit dates not readable on November 10, 2026.
How to Install Render MCP
$claude mcp add --transport http render https://mcp.render.com/mcp --header "Authorization: Bearer <YOUR_API_KEY>"Compatible with: Claude Code, Claude Desktop, Codex, Cursor.
Setup by client
This is a remote server. If it requires sign-in, your client opens a browser sign-in window (OAuth) the first time it connects. Clients without OAuth support cannot use servers that require it.
Cursor (~/.cursor/mcp.json)
{
"mcpServers": {
"render-mcp": {
"url": "https://mcp.render.com/mcp"
}
}
}VS Code (.vscode/mcp.json)
{
"servers": {
"render-mcp": {
"type": "http",
"url": "https://mcp.render.com/mcp"
}
}
}Claude Code (terminal)
claude mcp add --transport http render-mcp https://mcp.render.com/mcpAdd any API keys or environment variables the server's README lists.
About Render MCP
Render MCP is the Model Context Protocol server from Render (the cloud hosting platform) that lets an AI client create and manage Render services, databases, deploys, logs and metrics. Render runs a hosted version at https://mcp.render.com/mcp and says it strongly recommends that over running the server locally. A local Docker image and a Go binary also exist, but Render calls local running "not recommended". The server code is open source under Apache-2.0.
Quick facts (as of November 10, 2026)
| Render MCP | |
|---|---|
| Maker | Render (render-oss on GitHub) |
| Hosted endpoint | https://mcp.render.com/mcp (recommended by Render) |
| Local options | Docker image ghcr.io/render-oss/render-mcp-server, a prebuilt executable via install script, or build from source with Go |
| Auth | OAuth (browser flow) for supported clients, or a Render API key sent as a Bearer token |
| Tools | 25 listed in the repository README |
| License | Apache-2.0 (as listed on the repository page) |
| Price | The docs state no price for the MCP server. One metric needs a Pro workspace or higher |
| Repository status | No archive notice on the repository page. Latest GitHub release shown: v0.3.0 (14 January, year not shown on the page) |
Tools
The README groups its 25 tools by area:
- Workspaces:
list_workspaces,select_workspace(marked deprecated),get_selected_workspace - Services:
list_services,get_service,create_web_service,create_static_site,create_cron_job,update_environment_variables,update_web_service,update_static_site,update_cron_job - Deploys:
list_deploys,get_deploy,trigger_deploy - Logs:
list_logs,list_log_label_values - Metrics:
get_metrics - Render Postgres:
query_render_postgres,list_postgres_instances,get_postgres,create_postgres - Render Key Value:
list_key_value,get_key_value,create_key_value
The docs page describes the same capabilities in prose: create web services, static sites, cron jobs, Postgres and Key Value; trigger deploys (optionally clearing the build cache); read logs; read metrics such as CPU and memory, instance count, datastore connections, response counts by status code, response times and outbound bandwidth; and run read-only SQL against a Render Postgres database.
The README says autoDeploy defaults to yes and region defaults to oregon for new services, and that get_metrics accepts times only within the last 30 days.
Setup
Hosted server with OAuth
- Claude Code: the docs give a plugin route:
/plugin install render@claude-plugins-official, then/reload-plugins. - Claude Desktop: go to Customize, then Connectors, search for Render, click Connect and finish authorization in the browser.
- Codex:
codex mcp add render --url https://mcp.render.com/mcp --oauth-client-id codex - Cursor: run
/add-plugin renderin an agent window, choose a scope, add the plugin and click Authenticate.
Hosted server with an API key
For non-interactive setups, create a Render API key (the docs link to their API documentation for this) and send it as a Bearer token.
Claude Code:
claude mcp add --transport http render https://mcp.render.com/mcp --header "Authorization: Bearer <YOUR_API_KEY>"
Cursor (~/.cursor/mcp.json):
{
"mcpServers": {
"render": {
"url": "https://mcp.render.com/mcp",
"headers": { "Authorization": "Bearer <YOUR_API_KEY>" }
}
}
}
Codex (~/.codex/config.toml):
[mcp_servers.render]
url = "https://mcp.render.com/mcp"
http_headers = { Authorization = "Bearer <YOUR_API_KEY>" }
For Claude Desktop with a key, the docs use npx mcp-remote https://mcp.render.com/mcp --header "Authorization: Bearer ${RENDER_API_KEY}" with RENDER_API_KEY set in env. The docs add that other tools use different config keys, for example Zed uses context_servers and GitHub Copilot uses servers.
Local Docker (not recommended by Render)
The docs show a Docker config that runs ghcr.io/render-oss/render-mcp-server with -i --rm, passes RENDER_API_KEY, and mounts a volume named render-mcp-server-config at /config. An install script (macOS or Linux) and a build from source (git clone, then go build) are also documented.
Limitations and things to watch
- Destructive operations are possible. The docs warn that the server supports potentially destructive operations, such as changing environment variables and triggering deploys.
- Secrets. Render says the server attempts to minimize exposing sensitive information (like connection strings) to your AI app's context, but does not guarantee it. Use caution with secrets.
- Workspace scope. Authenticating grants access to your workspaces and services, and each action is scoped to the workspace you select. The README says the session-based workspace selection is deprecated and scheduled for removal, and that you should pass
workspaceIdon each call. - Limited service types and options. Only web services, static sites, cron jobs, Postgres and Key Value can be created. The docs say image-backed services and IP allowlists are not supported.
- Limited changes to existing resources. You can trigger deploys and modify environment variables. Other modifications and deletions need the Dashboard or REST API. The docs say scaling settings and other operational service controls cannot be changed through the server.
- Read-only SQL.
query_render_postgresruns read-only queries. - Plan requirement. The docs say web response time metrics require a Pro workspace or higher.
Pros and cons
Pros
- Official server from the platform vendor, with a hosted endpoint that needs no local install.
- OAuth for interactive clients, API key for non-interactive ones.
- Covers deploys, logs and metrics, which suit debugging a running service.
- Database queries are documented as read-only.
Cons
- Cannot delete resources or change scaling through the server.
- Only useful if your workloads run on Render.
- An API key grants access to your workspaces and services, so handle it as a secret.
- Local running is discouraged by Render, and the hosted server is the supported route.
Who it is for
A good fit if:
- You host on Render and want your assistant to read logs, check metrics, trigger deploys and query a Postgres database.
Not the right tool if:
- You host on another platform. See Vercel MCP, Netlify MCP or Heroku MCP.
- You need to delete resources or change scaling from the assistant. Those need the Render Dashboard or REST API.
What we could not verify
- Repository commit dates. The repository page showed no archive notice, but commit dates were not readable. The latest GitHub release is v0.3.0 dated 14 January, and the page does not show the year. Render's own changelog has entries about the MCP server dated July 17 and July 22, 2026, and we did not open the full entries.
- Exact OAuth scopes. The docs do not say what permissions an OAuth login or API key is limited to beyond workspace scope.
- Price of the server. Render's MCP docs state no charge for the server itself. We did not read Render's pricing pages.
- Behavior in practice. This page comes from reading documentation only.
How this page was put together
We read the official Render MCP documentation page, the repository README, the releases page and Render's changelog headlines on November 10, 2026, and compared the documented tools, authentication options and limits. We did not run the server.
Sources
Pricing
Open Source: from No MCP server charge stated in Render's docs (as of November 2026); web response time metrics need a Pro workspace or higher
Our Take on Render MCP
An official, vendor-run server for teams hosted on Render. It covers deploys, logs, metrics and read-only database queries, but cannot delete resources or change scaling, so it complements the Dashboard rather than replacing it.
Alternatives to Render MCP
Deploy, configure and debug Netlify sites through an AI agent
View Netlify MCP →