Agent-native
Tell your agent to ship it.
PrimDB speaks MCP. Claude, Cursor, and Codex deploy code, read your data, and hand back feedback through one endpoint at mcp.primdb.com. You describe the change. Your agent ships it.
One API
Deploy, data, and feedback from one endpoint
No four SDKs, no four dashboards. The surface your agent calls is the same platform you use by hand.
Deploy from a prompt
Your agent pushes a repo, opens a preview, and reports the live URL back in chat.
Read and change data
Query, migrate, and provision Postgres, Redis, search, and storage over the same API.
Pull feedback back in
Preview feedback returns to the agent that made the change, so a comment becomes the next iteration.
Setup
Add PrimDB to your MCP client
Drop this into your Claude Desktop or Cursor config, swap in a project token, and your agent can reach PrimDB.
{
"mcpServers": {
"primdb": {
"url": "https://mcp.primdb.com",
"headers": {
"Authorization": "Bearer <YOUR_PRIMDB_TOKEN>"
}
}
}
}Full setup and the tool reference land in the docs at primdb.com/docs/mcp.
Quickstart
Prompts to paste into your agent
Copy one of these, point your agent at the MCP endpoint, and watch it work.
“Deploy this repo to PrimDB and give me the preview URL.”
“Add a Postgres database to my PrimDB project and run the migration.”
“Show the open preview feedback on my latest PR and fix the top comment.”
Agent questions
Which agents can drive PrimDB?
Any MCP client. That includes Claude Desktop, Cursor, and Codex today, plus anything else that speaks the Model Context Protocol.
Is agent access safe?
Agents act with project-scoped tokens, and every action is audited. The MCP endpoint is the same platform you use by hand, so nothing runs in a side channel.
Where is the endpoint?
The Model Context Protocol endpoint is mcp.primdb.com. Point your MCP client at it with a project token to get started.
Point your agent at PrimDB.
One endpoint for deploy, data, and feedback. Actually free to start, no credit card.