all.sourceAllSourceEvent Store
Menu

Product demo

See event history become useful.

Move through one order history. Watch AllSource reconstruct past state, reveal its event timeline and graph, trace pipeline work, rebuild a projection, and answer through MCP.

Interactive capability lab

One event stream. Six ways to use it.

Follow one order from creation to dispatch. Move through its history and watch every AllSource surface answer from the same source events.

Entity

order-1042

Fixture

local · safe

Shared history cursor

As of 09:08:17 UTC · version 6

6 / 6 events applied

Source history

Event Timeline

Inspect every fact in sequence. Select one event to move every other view to that exact point.

Point-in-time read

Time travel without a second database

Replay source events through a timestamp. Current state and historical state come from the same durable stream.

status
dispatched
total
£129.00
payment
authorized
inventory
reserved
delivery_postcode
E1 6AN
shipment
ship_48 · Northline Express
Prime relationship view

Graph visualisation with provenance

Relationships appear only after their source event exists. Click a node to inspect what the graph knows at this point in history.

Order relationship graph as of event 6PLACEDPAID BYRESERVESSHIPS AS
Core inline processing

Trace one event through a pipeline

Filter, map, and branch accepted events without hiding which event produced each result.

Pipeline input

order.dispatched

  1. filter()

    entity_id = order-1042

    pass
  2. map()

    normalize payload

    done
  3. branch()

    route → orders

    orders
Pipeline transforms run in Core. Query Service projections below fold source history into tenant-scoped read models; they are a separate read-plane concern.
Query Service read model

Projection state you can rebuild

Fold the same ordered history into a query-ready shape, then keep it current as new events arrive.

order-summary ready · v6
status
dispatched
total
£129.00
payment
authorized
inventory
reserved
delivery_postcode
E1 6AN
shipment
ship_48 · Northline Express
6 events folded · HTTP, realtime, or analytics read paths
Agent tool interface

MCP data access agents can reason over

Use explicit, tenant-scoped tools instead of pasting whole histories into prompts. These are real event-store connector tool shapes.

Choose an MCP tool
tools/callrequest
{
  "name": "event_timeline",
  "arguments": {
    "entity_id": "order-1042",
    "until": "2026-08-14T09:08:17Z"
  }
}
contentsuccess
{
  "entity_id": "order-1042",
  "count": 6,
  "timeline": [
    {
      "version": 1,
      "timestamp": "2026-08-14T09:00:04Z",
      "event_type": "order.created",
      "summary": "Order opened for customer cus_mara"
    },
    {
      "version": 2,
      "timestamp": "2026-08-14T09:01:18Z",
      "event_type": "order.confirmed",
      "summary": "Customer confirmed checkout"
    },
    {
      "version": 3,
      "timestamp": "2026-08-14T09:02:41Z",
      "event_type": "payment.authorized",
      "summary": "£129.00 payment authorized"
    },
    {
      "version": 4,
      "timestamp": "2026-08-14T09:04:09Z",
      "event_type": "inventory.reserved",
      "summary": "Two items reserved in warehouse LON-2"
    },
    {
      "version": 5,
      "timestamp": "2026-08-14T09:05:32Z",
      "event_type": "delivery.address_corrected",
      "summary": "Delivery postcode corrected before dispatch"
    },
    {
      "version": 6,
      "timestamp": "2026-08-14T09:08:17Z",
      "event_type": "order.dispatched",
      "summary": "Parcel handed to Northline Express"
    }
  ]
}
Interactive fixture · no network request · connector-compatible tool names and parameters

Durable source

Core keeps the ordered events in WAL + Parquet.

Separate reads

Query Service serves HTTP, realtime, analytics, and projections.

Agent interface

MCP tools expose history and state without bulk prompt stuffing.

Build from working code

Demo explains the model. These examples show real SDK and service integration paths.

Reference projects

Real services in the monorepo that consume AllSource as their data layer. Read the source to see how the pieces fit together end-to-end.

Solution walkthroughs

Long-form pages on how AllSource maps to specific problem shapes — closer to a tutorial than a snippet.