OBJECTSTACK · AI-NATIVE APPLICATION PLATFORM

Low-code does not remove code. It turns code into an operable business model.

ObjectStack connects development and operations through protocol-defined metadata. Developers define objects, views, flows, permissions, and integrations precisely; teams configure, execute, and analyze in Studio and Console. The Bitdata AI plugin brings governed agents into the same runtime.

See code become an application Explore AiCRM

CODE-DEFINED LOW-CODE

Define the application in code. Keep low-code speed for daily work.

TypeScript metadata describes objects, fields, relationships, views, pages, navigation, actions, flows, approvals, permissions, reports, and translations. ObjectStack validates the build and renders it in Studio and Console.

Protocol schemas are executable contracts
Review every business-model change in Git
One metadata model drives UI, API, and automation

DATA WITHOUT SILOS

Bring external data in—and bring it under business governance.

Native CSV/XLSX import provides mapping, server preflight, background jobs, failure evidence, and bounded undo. REST, OpenAPI, and domain connectors handle continuing synchronization. Permissions, hooks, validation, and audit apply as data enters objects.

Background import path up to 50,000 rows
Failed rows and job history remain traceable
Every write passes object-level governance

MCP-NATIVE OPERATIONS

MCP connects real business services, not just chat demos.

The ObjectStack MCP connector brings trusted services into the automation and action model. Hosts explicitly allow transports and commands while execution continues to honor identity, permissions, RLS, approval, and audit.

MCP, REST, and OpenAPI share a connector model
stdio is denied until explicitly trusted by the host
Model prose cannot turn a connector failure into success

IN-APP AGENTS

Agents converse inside the application—and stay within its rules.

@bitdata/ai-plugin provides an installable streaming runtime, Ask Agent, skills, tools, knowledge retrieval, and approval-aware actions. Extensions use public registries and the ai:ready hook instead of private source imports.

Explicit Agent → Skill → Tool composition
Caller identity and environment follow tool execution
High-risk actions enter human approval

Metadata is product source

One definition, many runtime surfaces.

Application packages depend only on public ObjectStack contracts. Once validated, the metadata drives database objects, REST resources, Console navigation, business actions, flows, and the AI tool catalog.

01Studio schema02Console workspace03REST resources04Automation & approvals05Agent tool catalog

objectstack.config.tscrm.object.tsquote-approval.flow.ts

export default defineStack({
  plugins: [new BitdataAIPlugin()],
  objects: allObjects,
  views: allViews,
  actions: allActions,
  flows: allFlows,
  permissions: allPermissionSets,
  apps: Object.values(apps),
});

Schema validatedobjectstack build

Rules apply the moment external data enters

From files or services to trusted business objects.

Source

CSV · XLSX · REST · OpenAPI · MCP

Shape

Map · type · relate · preflight

Govern

Hook · validate · permission · RLS

Operate

View · flow · report · agent

The intelligence layer does not create a second authorization system

Models reason. The business runtime owns the facts.

Agents see only skills and tools registered in the active kernel and available to the caller. Reads come from permission-aware queries; actions come from real tool results. Risky operations re-check authorization and target state after approval.

AgentSkillTool

Ask AgentObjectStack business context

LIVE

Analyze quote discounts and win rate over the last 30 days, then explain the outlier accounts.

schema_readerdata_exploreranalytics_query

Current environment, user permissions, and object row policies applied

Quotes discounted over 15% gained 6.2 points in win rate while margin fell 4.8 points. Five outlier accounts were identified; no write operation was executed.

TRUST LAYER

The same person exists from page click to tool execution.

Identity and environment

Authenticated user, positions, and environment follow data, knowledge, automation, and action calls.

Least privilege

Missing trusted identity fails closed; available tools change with kernel services and caller permissions.

Human approval

Destructive or high-risk actions are not directly executable, and approval never bypasses underlying access checks.

Audit and real outcomes

Model prose is not success. Only tool results and business audit facts confirm that work completed.

BUILT ON OBJECTSTACK

Choose a business domain and see the application.

AiCRM serves general B2B lead to cash. IP CRM serves professional delivery, matters, deadlines, and rights assets.

Explore AiCRM Explore IP CRM