Render
by Render
Create services, provision Postgres and Key Value stores, trigger deploys, query databases and read logs from your AI assistant. 25 tools, OAuth sign-in, and no tool that can delete anything.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Render via MCP
https://mcp.render.com/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Render Tools & Capabilities (25)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- No deletion, suspension, restart or scaling. Render's MCP documentation directs you to the Dashboard or REST API for anything beyond deploys and environment variables.
- No rollback tool. Rollbacks exist on Render but not in this connector.
- Only five resource types can be created: web services, static sites, cron jobs, Render Postgres and Render Key Value. Render states other service types are not yet supported.
- Not all creation options are supported. Render names image-backed services and IP allowlists as examples you cannot configure through the server, and
create_web_servicecannot create Docker or container-registry services. - Three
update_*tools are no-ops that return a dashboard link. - No scope narrowing. Neither discovery document publishes a
scopes_supportedkey. - No read-only mode and no per-service restriction. Scoping is per workspace only.
- Response-time metrics require a Pro workspace or higher.
- No spend confirmation. Approval in your MCP client is the only gate before provisioning.
- Secrets are minimised, not guaranteed. Render states the server attempts to avoid exposing sensitive information such as connection strings to your AI app's context but does not guarantee it.
- Rate limits apply to the REST API behind the tools. Render publishes 20 service creations per hour, 10 deploy or update calls per minute per service, 30 log requests per minute, and 400 other GET requests per minute, returning HTTP 429 with
Ratelimit-RemainingandRatelimit-Resetheaders. select_workspaceis deprecated and scheduled for removal.- The server does not query Render's documentation. Render notes its primary MCP server has no docs tools and points at a separate experimental documentation server, which it may discontinue at any time.
Frequently asked questions
Can the Render MCP server delete a service or a database?
No. None of the 25 tools deletes, suspends or destroys anything. Render's REST API publishes Delete service, Delete Postgres instance, Suspend service and Restart service endpoints, and its MCP documentation states that for modifications and deletions beyond deploys and environment variables you must use the Render Dashboard or the REST API instead.
Will an AI agent using the Render MCP server create infrastructure that costs money?
Four of the five create tools default to a free plan and one does not. Render's published server source defaults create_web_service, create_postgres and create_key_value to free, and static sites are free to deploy. Cron jobs have no free instance type, so create_cron_job defaults to Starter at 0.00016 dollars per minute of runtime.
Can the Render MCP server write to your Postgres database?
No. The query_render_postgres tool runs SQL inside a read-only Postgres transaction, which Render states in the tool's own description and implements in the published server source by opening the transaction in read-only access mode. A write statement fails at the database rather than being filtered by the server.
Can an agent roll back a Render deploy it broke?
Not through the MCP server. Render supports rollbacks and its REST API publishes a Roll back deploy endpoint, but no rollback tool exists among the 25. The recovery path is the Render Dashboard or the REST API. Render notes that a dashboard rollback also disables automatic deploys, while an API rollback does not.
What does update_web_service actually do on the Render MCP server?
It returns a dashboard link rather than changing anything. Render's server README states that direct updates to a web service are not supported through the MCP server and that the tool returns a link to the service's settings. The same is true of update_static_site and update_cron_job. Only update_environment_variables writes.
Does update_environment_variables wipe environment variables the agent did not pass?
Not by default. Render's published server source documents that supplied variables merge with the service's existing ones, so the agent never needs to read your secrets into its context. Setting the replace parameter to true instead replaces the whole set, which deletes any variable the agent omitted.
Which OAuth scopes does the Render MCP server request?
None are published. We fetched Render's OAuth protected-resource descriptor and its authorization server metadata on 2026-08-18, and neither document carries a scopes_supported key at all. There is no narrower grant to choose, so the consent screen is the only place the connection's breadth is described.
Can you connect to the Render MCP server with an API key instead of OAuth?
Yes. Render's MCP documentation gives API key setup alongside OAuth for Claude Code, Claude Desktop, Codex and Cursor. Note that a Render API key carries your full REST API reach, including the delete and suspend endpoints the MCP tools do not expose, so an OAuth sign-in is the narrower of the two credentials.
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Developer Tools
- Developer
- Render
- Tools
- 25
- Domain
- mcp.render.com
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.