Your agent builds.Your data stays local.

Localbase gives Codex a scoped MCP surface for creating auth, resources, relationships, indexes, and rows against a real local Postgres backend.

1

Get started

No hosted account required. Install locally and connect your agent in minutes.

Start a Localbase workspace

Docker, Postgres, API config, and MCP setup from your terminal.

Install Localbase
2

Connect your agent

Install the scoped MCP server so Codex can make backend changes through Localbase.

> npx @mrace07/localbase init ai-memory
OK Localbase project created
OK MCP server ready on localhost
3

Give it tasks

Approve backend changes, inspect generated resources, and keep the data on your machine.

APPROVE

Create auth-owned memories and documents

resources.create / indexes.add / rows.insert

How it works

1Give your agent a task
2Approve the change
3Every request is typed

"Create a backend for an AI research assistant." Codex figures out the resources it needs and calls the Localbase MCP tools.

Create memories, documents, conversations, citations, and saved outputs.
I will need access to:create_resourcecreate_relationshipinsert_row
+

Risk Assessment

Codex gets typed backend operations instead of broad shell access or improvised SQL.

+

Credential Injection

Project config stays local while the agent works through a scoped MCP server.

+

Audit Trail

Resources, rows, indexes, auth, and relationships are inspectable in the generated backend.

Self-hosted by default

The demo video lives in the app assets folder and is imported into this UI. It shows the local workflow instead of relying on an external embed.

14 backend tools and counting

Every operation gets a narrow interface, local state, and a repeatable command path.

PostgresAuthResourcesIndexesRelationshipsSeed rowsMCPSDKDockerCLICodexREST APILocal usersAudit logs

Localbase for teams

Managed local development with reproducible setup, visible schema changes, and project-level agent access.

+

Private local runtime

+

Project-scoped MCP

+

Postgres data model

+

CLI-first workflow

Contact us