Skip to directory

About Prisma MCP

Prisma's hosted MCP server lets AI clients manage Prisma Postgres, Prisma Compute and Object Storage workspaces, including database operations, SQL queries, backup restores, logs, deployments and access keys.

What you can use it for

  • List, create and manage Prisma Postgres databases
  • Run SQL, inspect schemas and restore backups
  • Inspect and manage existing Prisma Compute deployments, read logs and set environment variables
  • Manage Object Storage buckets and access keys
  • Inspect Composer service topology and search Prisma docs
Connection and setup
Hosting
Remote, vendor-hosted
Transport
Streamable HTTP
Endpoint
https://mcp.prisma.io/mcp
Authentication
OAuth sign-in with a Prisma account. On connection, choose a workspace and approve access. Access follows that account/workspace; no API key is required in the documented remote setup.

Requirements

  • Prisma account
  • Access to a Prisma workspace and resources
  • MCP-compatible client with remote Streamable HTTP and OAuth support

Setup steps

  1. Add https://mcp.prisma.io/mcp as a remote Streamable HTTP server in the AI client.
  2. Connect and sign in to Prisma; select the intended workspace and approve access.
  3. Verify using a read-only request such as listing databases. Check the target workspace before any operation that creates, changes or deletes resources.
Tools 31

The official Prisma MCP tools reference enumerates all 31 stored identifiers. Exact documented list; live access still depends on workspace and credentials.

  • fetch_workspace_details
    01
  • list_prisma_postgres_databases
    02
  • create_prisma_postgres_database
    03
  • delete_prisma_postgres_database
    04
  • list_prisma_postgres_connection_strings
    05
  • create_prisma_postgres_connection_string
    06
  • delete_prisma_postgres_connection_string
    07
  • list_prisma_postgres_backups
    08
  • create_prisma_postgres_recovery
    09
  • introspect_database_schema
    10
  • execute_sql_query
    11
  • execute_prisma_postgres_schema_update
    12
  • list_object_store_buckets
    13
  • create_object_store_bucket
    14
  • delete_object_store_bucket
    15
  • create_object_store_bucket_key
    16
  • delete_object_store_bucket_key
    17
  • search_prisma_documentation
    18
  • list_prisma_compute_apps
    19
  • list_prisma_compute_builds
    20
  • list_prisma_compute_deployments
    21
  • get_prisma_compute_deployment_logs
    22
  • promote_prisma_compute_deployment
    23
  • rollback_prisma_compute_app
    24
  • start_prisma_compute_deployment
    25
  • stop_prisma_compute_deployment
    26
  • delete_prisma_compute_deployment
    27
  • delete_prisma_compute_app
    28
  • set_prisma_compute_env_var
    29
  • delete_prisma_compute_env_var
    30
  • get_prisma_composer_topology
    31
Limitations and notes
  • MCP operations that change or delete resources may cause data loss or interrupt live traffic; the AI client's approval settings should be reviewed.
  • Prisma states that deploying new source code is handled through its separate Prisma Compute guide, not by assuming a single MCP deploy tool.
  • Requires access to a Prisma workspace; resources outside the signed-in user's permissions are not available.
Source discrepancies
  • ?Prisma Postgres? discovery entry is resolved to the broader official Prisma MCP server, which covers multiple Prisma services. It is not limited to a local Prisma CLI server.
Original sources
  • official tool reference

    Official Prisma MCP reference enumerates the 31 stored tools across workspace/database management, connection strings, backups/recovery, schema/query, object storage, docs search, compute deployments and environment variables.

  • official documentation or source

    Official Prisma MCP reference enumerates the 31 stored tools across workspace/database management, connection strings, backups/recovery, schema/query, object storage, docs search, compute deployments and environment variables.

Documentation checked 2026-10-28T12:53:18.000Z. Connection has not been runtime tested.