Sign in Start free trial
Platform

Integration

Connect any application: a public API, events, webhooks, connectors, MCP and a data workbench.

 AI gives every hub a ready-made public API with scoped keys, documented events and signed webhooks, plus a data integration module with connectors, pipelines and an on-premises agent. Tools that speak the Model Context Protocol can read and prepare work as a signed-in person. Nothing an integration does can exceed the rights of the key or person behind it.

54 features available 7 planned 7 capability areas
Highlights

What stands out

Every hub, one API

A single versioned API with scoped keys gives integrators one thing to learn.

Signed events you can trust

Documented events and signed webhooks replace polling.

Data workbench

Design, test and monitor pipelines in one workspace with lineage.

Works behind your firewall

An on-premises agent reads local data without opening inbound ports.

AI tools as a person, never more

MCP clients inherit a person's rights and cannot approve.

Capabilities

Every feature, in one list

Search or filter. Planned items are labelled; everything else is available now.

Public API

One versioned REST gateway over a shared catalogue of business objects.

14 features
  • Versioned /v1 REST gatewayOne gateway for every hub, with cursor paging, idempotency keys, ETags and rate limits.
  • API catalogue in the appBrowse every resource, its fields, events and coverage by hub.
  • OpenAPI per tenantA generated OpenAPI document that includes the tenant's own custom fields and hides unlicensed resources.
  • Scoped API keys and client credentialsKeys and OAuth client credentials limited to named permissions.
  • Usage and limits per keyDaily usage, rate overrides and monthly quotas per key.
  • Batch callsSubmit many operations in one call, atomic or not.
  • Change feedAsk for everything updated since a point in time.
  • REST actionsCall business commands such as converting or submitting; person-only decisions are refused for keys.
  • Approval-routed writesA write that needs approval returns accepted and goes to workflow.
  • Bitemporal version routesRead records as they were at a date.
  • File import through the same pathBulk files are validated by the same rules as the API.
  • Wider table coveragePlannedMany tables are not yet reachable over /v1; coverage is being widened hub by hub.
  • Developer portal with sandbox keysPlannedQuickstarts, sample collections and self-service test keys.
  • Dated API versions with deprecation headersPlannedA published deprecation policy and migration notes.

Events and webhooks

Know when something changes, without polling.

7 features
  • Documented event catalogueEvery subscribable event is documented and guarded by tests.
  • Generated change eventsEvery writable resource announces a change event from its hub with ids only.
  • Signed webhooksDeliveries are signed so receivers can verify them, with retries and replay protection.
  • Safe outbound callsWebhooks and connectors share one guard against server-side request forgery.
  • Inbound endpointsReceive signed messages from other systems, map them and keep an error list.
  • Real-time incentive eventsA separate ingest API for sales events with lookups, evaluations, wallets and estimates.
  • Durable outbound event streamPlannedAn ordered, replayable feed for consumers that want a stream rather than webhooks.

Developer tools

Build against  AI from your own tools.

7 features
  • TypeScript SDKA typed client covering every /v1 resource, with webhook signature checks.
  • Command-line toolLog in with profiles, pull and push configuration, and run script tests in CI.
  • Configuration as codePlan and apply configuration from files, with a protected production profile.
  • Local test harness for extensionsRun extension scripts against saved fixtures on your own machine.
  • Sandboxed scriptsTenant scripts run in an isolated runtime with time and memory limits and no network by default.
  • Embed kit for dashboardsSigned tokens and a JavaScript kit to embed analytics.
  • Generated clients in other languagesPlannedClients generated from the OpenAPI for other languages.

Custom APIs

Publish your own endpoints without writing a server.

5 features
  • Custom report endpointsExpose a saved report as an endpoint with parameters and an output mapping, run as the caller.
  • Custom workflow endpointsAccept a validated payload and run an approved action such as an upsert through approvals.
  • Immutable published versionsPublish, retire and dry-run test custom endpoints.
  • Declarative mapperMap fields between systems with pointers and fixed transforms.
  • External id linksKeep a link between your record and the other system's id.

Connectors and data pipelines

Move data in and out with a visual designer and a library of connectors.

20 features
  • Visual pipeline canvasDesign flows on a canvas or in code over one definition, with versions, diffs, dry runs and run history.
  • Data workbenchAn IDE-style workspace with explorer, catalogue, palette, shortcuts and an assistant pane.
  • Data catalogue and lineageDatasets with owners, tags, quality score, lineage and glossary.
  • Data quality rulesRules with block severity that stop a load, and a score on each dataset.
  • Restart from the failing stepResume a failed flow without repeating finished steps.
  • Orchestration and monitoringDashboards for runs, retries and dependencies.
  • On-premises agent fleetSigned instructions, key rollover and fleet screen for reading data behind your firewall.
  • Streaming sourcesWebhook and database streaming inputs.
  • API workbenchExplore APIs with environments, saved collections and masked history.
  • Explain and optimise SQLExplain queries and suggest improvements; changes to database objects go through approval.
  • Matching and transformsMatch records across sources and transform fields in flows.
  • File sourcesUpload, S3, Google Cloud Storage, Azure storage and SFTP files.
  • Database sourcesRelational databases with tables and queries, through the cloud or the agent.
  • Warehouse sourcesSnowflake, BigQuery, Databricks, Redshift, Synapse and Fabric style warehouses.
  • Business application sourcesCommon CRM, ERP and accounting systems through their own connectors.
  • Enterprise application sources
  • Knowledge sourcesPull help and knowledge articles from document and wiki systems.
  • EDI and SOAP exchangeBusiness documents over EDI and SOAP for order exchange.
  • Kafka streaming sourcePlannedBuilt, available in test environments only for now.
  • Live validation of vendor connectorsPlannedConnectors are tested against recorded replies; live vendor accounts are not yet validated.

MCP and agents

Let approved AI tools work inside  AI as a person.

5 features
  • MCP serverTools generated from the same intent registry the assistant uses, over a standard transport.
  • OAuth for AI clientsConnect a client with an authorisation a person grants and can revoke.
  • Rate limits per clientA token bucket limits each client.
  • Prepare, then a person authorisesAgents can prepare; money and sensitive actions need a person's authorisation.
  • MCP administrationAdmins manage clients and what tools they see.

Application add-ins and chat

Reach people where they work.

3 features
  • Word and Outlook add-ins for contractsFindings beside clauses and filing of e-mails, signed in with a person's approval.
  • Document management connectorsSharePoint, Google Drive and Box adapters for filing and import.
  • Service-management discovery connectorsDiscovery and event connectors for IT service hubs.
Assistant

What  AI prepares here

It prepares the work. A person confirms it.

  • Draft a pipeline or flow from a description
  • Describe a dataset in the catalogue
  • Run a trial of a flow
  • Find an asset in the workbench
  • Search and explain a connector pack
  • Draft a custom endpoint for a person to publish
  • Save a script draft (publishing stays with a person)
Inside every hub

Workflow, reporting, integration, migration

The same four building blocks, specific to Integration.

Workflow

Publishing custom endpoints, certifying datasets and changing database objects can route through workflow approval.

  • Database object changes in the workbench need approval
  • Custom endpoint publication needs a second person
  • A write over the API that needs approval returns accepted

Reporting and analytics

Usage per key and delivery logs are visible to admins, and run history feeds monitoring.

Datasets

API usage per keyPipeline runsData quality scores

Dashboards

Orchestration and monitoring

Integration

This page is the integration module itself; each hub shows its own Integration sub-hub.

/v1 with several hundred named resources; coverage differs by hub

Events

Resource changed events per hubDocumented hub eventsApproval eventsAutomation events

Connectors

Files and object storageDatabasesWarehousesBusiness applicationsEnterprise applicationsEDI and SOAPKnowledge sourcesMCP

Migration

Connector packs from the migration module reuse the same engine.

Mobile

On the phone

Not applicable; admin screens are desktop first.

Who uses it

Integration developerIntegration administratorIntegration operationsIT securityPartner or ISVData engineer
Questions

Frequently asked

Does every hub have an API?

Yes, through one gateway. Coverage differs: some hubs are fully readable and most writable, others are thinner and widening.

Can an integration approve something?

No. Approvals, rejections and similar decisions are person-only.

How do I get notified of changes?

Subscribe to documented events by webhook, or poll the change feed.

Can I read data behind a firewall?

Yes, with the on-premises agent.

Can AI tools connect?

Yes, over MCP with a person's authorisation, limited to that person's rights.

Is there an SDK?

A TypeScript SDK and a command-line tool exist; other languages are planned.

Can I expose my own endpoint?

Yes, from a saved report or a validated workflow action, versioned and tested before publish.

See Integration on  AI

Talk to us