Open District
MENU
What it is Agents Governance Status Access
Request early access

All data. One memory.
One truth.

Open District keeps everything your people and your agents learn — public data, enterprise systems, personal context, sensors, news — in one memory you can query and trust. Every fact has a source. Every answer has citations.

Request early access Connect an agent
live evidence · checking…
probing real endpoints…

Not a mockup. This card probes the running system and shows what it answered, verbatim. raw JSON

What it is

One memory. Governed.

Open District is a governed memory layer: an append-only record of everything you connect, queryable in SQL or natural language, where facts keep their history — when something became true, what changed it, and who observed it. People ask it. Teams trust it. Agents build on it.

  • Public data
  • Enterprise systems
  • Personal context
  • Sensors
  • News
  • Agent activity

For people

Ask in plain language. Get answers with citations you can check — not confident guesses.

For teams

One governed record instead of ten systems that never reconcile. Isolated by tenant, logged by default.

For agents

A persistent, governed memory over MCP — so your agents remember between sessions and cite their sources.

Proof

No invented customers. No invented numbers.

There is nothing here yet — and that is the point.

We don't show logos we don't have or metrics we made up. When real customers and real answers land, they will appear in this exact spot — served from our own memory, with citations. The page grows as the platform grows.

For agents

Your agents already know how to use it.

Open District speaks MCP. Point any agent at one endpoint and it can query, search, remember, and reason over your governed context — under an API key with explicit scopes, audited like every human query.

{
  "mcpServers": {
    "opendistrict": {
      "url": "https://mcp.opendistrict.org/mcp",
      "headers": {
        "Authorization": "Bearer $OD_API_KEY"
      }
    }
  }
}
Read agent setup

Streamable HTTP transport. Keys are issued per tenant with explicit scopes. Requests are governed and audited exactly like the web app. One-click client adds (Cursor, Claude Desktop, Cline) are on the roadmap.

Machine-readable, by design

This page ships /llms.txt — capabilities, endpoint, status — so agents can onboard without a human in the loop.

10 MCP tools today: query, search, schema, catalog, topology, metrics. All read-governed, all audit-logged.

Available now

For organizations

Built for governed action.

Every query, answer, memory, and action is scoped, isolated, and auditable.

Tenant isolation Available now

Database-level row-level security, not app-level filtering. Enforced by Postgres RLS, fail-closed.

Scoped API keys Available now

Keys carry explicit scopes. No ambient privilege.

Read-only query gates Available now

Queries pass an AST-level read-only gate and table allowlist before execution. Injection is not possible.

Append-only audit log Available now

Every query and result recorded with actor, scope, and timestamps — queryable like any other table.

SSO / OIDC · PASETO sessions Available now

OIDC sign-in; PASETO session tokens with denylist revocation.

Human-in-the-loop approvals In development

High-impact actions require human sign-off before execution. Gate points defined; workflow surface in development.

Enterprise-grade controls. Compliance roadmap available on request. No certification badges we don't hold — that is a feature, not a gap.

Live status

The whole vision, honestly labeled.

Everything we intend to ship, and exactly where each piece stands today. Chips flip as features land — the page states the destination from day one and never has to lie about the present.

Understand

Available now

Ask in SQL or natural language over every connected source. Federated queries span the OLTP and OLAP tiers; semantic search covers the full history.

Remember

Available now

Append-only event store: every record kept with source, timestamp, and payload. The basis for citation-backed answers.

Reason

Available now

Entity graphs connect records across systems — every hop cited to the event it came from.

Control

Available now

Row-level isolation, scoped keys, read-only gates, and an append-only audit log of what was asked and answered.

Act

In development

Governed actions and approvals turn insight into execution. Scheduled pipeline jobs run today; the human-approval loop is in development.

Predict

Platform direction

Forecasting and anomaly detection over stored context. Not built yet — labeled plainly rather than implied.

Next on the horizon: network-wide intelligence — aggregate signals across entities without exposing raw tenant data. Labeled direction until it exists, because privacy trust is the product.

Available now In development Platform direction

Nothing on this page claims a capability the platform does not have. Status chips are verified against the running system, and the evidence above is fetched live — check it yourself.

Get access

Give your organization a memory.

Open District is in early access. Tell us what you want to build — we read every submission.