Ingest
200+ connectors
ERP, CRM, ITSM, document stores, warehouses, and the long tail.
The 400 systems that don't talk.
Connect replaces the CSV ritual with a nervous system that actually works — and gives your AI safe hands, not just eyes.
What is SphereIQ Connect?
SphereIQ Connect is the platform's integration fabric: 200+ enterprise connectors, change-data-capture, event streaming, and an API gateway — plus an MCP hub that exposes every connected system as a governed tool AI agents can read from and act on.
In, through, and — this is the part that matters — back out.
Move
CDC & streaming
Change-data-capture and an event bus keep the platform current in seconds, not nightly batches.
Real-time CDCEvent streamingBatch ETL / ELTWebhooks & API pull
Act
MCP hub & reverse actions
Every system becomes a governed MCP tool — so agents can create the invoice, update the ticket, file the record. With policy and approvals enforced.
MCP servers per connectorScoped, governed actionsHuman-approval gatesFull audit trail
How a sync actually works.
Five steps between a change in a source system and a record your AI can safely use — the same path, every connector.
A webhook or schedule fires.
The event lands in a durable change-data-capture queue.
Durable, Postgres-backed CDC queue
Bursts are coalesced into one incremental delta sync.
Burst coalescingIncremental sync cursorsFull vs delta replace
Content is chunked and embedded.
Secrets are redacted at ingest and source permissions are captured with every record.
Mandatory redactionACL principal capture
From the connector catalog.
SharePointConfluenceNotionGoogle DriveOneDriveDropboxFile uploadBoxAmazon S3SlackMicrosoft Teams
ZendeskFreshdeskIntercomSalesforceNetSuiteHubSpotGitHubGuruGitLabJiraWeb crawl
The developer surface.
For the systems the catalog doesn't cover — and the teams that build on the fabric itself.
Connector SDK
Build
A connector implements one collect() method — the runtime handles chunking, embedding, redaction, permissions, and incremental sync.
MCP endpoint
Integrate
A JSON-RPC endpoint exposes governed recall and remember tools to any MCP-capable client — agents built anywhere can use it.
Outbound webhooks
Extend
Webhooks for message, conversation, and budget events. Auth models: OAuth, API key, upload, and crawl.
An assistant reads your systems. A workforce acts on them.
Most AI deployments stop at retrieval — answers about your business. Connect's reverse-action layer is what turns answers into outcomes: a NetSuite invoice created, a Monday.com item updated, a ServiceNow ticket resolved. Every action policy-checked by Governance before it executes, and written to the audit ledger after.
200+
Enterprise connectors, prioritized by real client stacks
Days, not quarters
Typical time to first live connector set in a deployment
100%
Of agent actions pass through the Governance plane. No exceptions.
The Integration Readiness Checklist
The 22 questions our architects ask before any deployment: which systems, which auth models, which data owners, which quiet CSV rituals. Run it internally and you'll know exactly where your integration debt lives.
Asked before every integration.
No. Connect coexists with existing integration platforms — it can consume their outputs and events. What it adds is the AI-specific layer they don't have: MCP tool exposure, governed agent actions, and semantic routing into the Company Brain.
Three paths: the generic API/webhook connector, a Sphere-built custom connector (typically 1–3 weeks), or certified partner-built connectors. The catalog is prioritized by real customer stacks, so the common cases are already covered.
Every action is a scoped MCP tool call that passes through the Governance plane: policy check, approval gate if the policy requires one, execution, then an immutable audit record. Agents never hold raw credentials — scopes are granted per agent, per tool, per policy.
Once the systems talk, your company can finally remember.
Chapter 02 is where the connected data becomes a mind.