Live dashboards
Admin panels, dispatch boards, and metrics that re-render the moment data changes—queries push to the screen; sync pulls in everyone's edits.
Build live tools people and agents run together
GoatDB gives browsers and agent runtimes local, reactive state that synchronizes through one coordinating server—so you can build the workflow instead of assembling its shared-state platform.
field-report-07you flaggedfield-report-07agent draftingfield-report-07you confirmed
What becomes practical
Build the tool your work needs
Dashboards, approvals, rosters, and field apps—many tools over one shared state 
, live for people and agents alike.
Approval queues
Requests with statuses, owners, and an attributable record of who approved what—teammate or agent.
Shift boards
One live roster the whole crew checks from their phones; swaps route through a manager's approval and reach every device moments later.
Field apps that work offline
Inspections, job capture, and sign-offs with zero signal; the phone holds the truth and syncs when coverage returns.
Offline-first utilities
Checklists, budgets, and trackers that open instantly, keep working offline, and sync when connected.
Agent teams
Several agents divide the work, each with its own permissions, every action attributable; people steer and approve from any device.
Architecture compression
Years of platform work, collapsed into the application
Realtime collaboration usually means stitching together storage, APIs, offline replay, synchronization, merge, provenance, recovery, and operations. GoatDB integrates that shared-state burden into a single deployable application/server binary.
The conventional platform project
Specialized systems to build, connect, and operate
- DatabasePersistence and indexes
- API layerAccess and transport
- Realtime syncFan-out and replay
- Offline stateLocal reads and writes
- Merge logicConcurrent changes
- Audit historyAttributable mutations
- RecoveryHistory and restoration
- OperationsDeploy and maintain
The GoatDB path
One deployable shared-state system
Local repositories, reactive queries, authorization, sync, merge, signed history, persistence, and bounded replica-assisted recovery.
The application/server plane is packaged together. Browser clients and external model or tool services remain separate.You still own the application
- ExperienceUI, domain model, and product behavior
- IdentityMap real actors to application sessions
- AgentsOrchestration, tools, retries, and guardrails
- PolicyApproval semantics and human checkpoints
GoatDB owns the shared-state mechanics
- Local statePersistence and reactive in-process queries
- SynchronizationAuthorized server-coordinated repository sync
- HistoryStructural merge and signed secure-mode commits
- Shared-state serverAccess checks, persistence, and repository isolation
Measured speed
One loader. Then nothing ever loads
Apps where nothing ever spins are usually out of reach—they take an architecture most teams can't afford to build. GoatDB makes it the default: each repository loads once, the way a desktop app opens a document, and from then on every read, write, and screen update runs at memory speed—the same in the browser as on the server. No per-interaction spinners, no optimistic-state machinery, no cache tier to maintain.
300×faster reads than SQLite in the browser7× on the server. Measured, same machine, reproducible.
1.5µsreads
7µswrites
<0.1µslive query results
~650ms, onceopening 100k items
A strong fit when
- Bounded working setsOperational data that divides cleanly into repositories.
- Repeated collaborationPeople and agents revisiting and changing shared state.
- Offline-capable workUseful local interaction before remote convergence.
- Attributable changesSecure-mode histories tied to signing sessions.
Design around the boundaries
- Each opened repository is a full local replica loaded into memory, so keep repository boundaries bounded.
- Cold predicates scan the working set; GoatDB is not SQL or an analytics engine.
- Remote convergence is polling-based—typically 700–1000ms—not cursor-grade transport.
- Secure writes pay per-commit signing cost; measure bulk-ingest workloads.
Pair GoatDB with PostgreSQL or an analytics store for relational queries, warehousing, and unbounded datasets.
Open source · MIT licensed
Start with shared state, then shape the app around your work
Initialize a project, connect the people and agents your workflow needs, and keep the whole shared-state plane in code you can inspect.
npx -y @goatdb/goatdb init