Integrations & Atoms

Every system you run — including the ones behind your firewall

Typed connectors with credentials kept out of sight, SOAP and JDBC for the systems that predate REST, message brokers in, webhooks out, connectors that write themselves — and Atoms, a hybrid runtime that executes flows on your own premises and returns only the result.

Connectors

Typed operations, not hand-rolled HTTP

Install a connector from the Marketplace, store credentials in an api Connection, and call its operations from the connector step with typed inputs and mapped outputs. The step-config UI walks you from connector to operation to inputs to connection — no URLs, no headers, no secrets in variables.

  • Declarative descriptors — a connector is a versioned contract of operations, auth, defaults, and fixtures, validated at publish time
  • Auth handled — API key, basic, bearer, and OAuth2 client flows, with a hosted consent screen for user-authorised connections
  • Pinned, SSRF-safe egress — one DNS resolve, address validation, socket pinned to the validated IP, TLS SNI against the original host
  • Declared egress domains — a call outside a connector's allowlist is flagged by audit_app
  • Account-local connectors for private integrations you never publish
  • Long-running vendor jobs: async_job_drain creates, polls to completion on durable intervals, and drains paged results
 Step: connector
🔌 Connector: hubspot v2.3Installed
⚙️ Operation: contacts.upsertTyped
🔑 Connection: hubspot-prod (OAuth2)Secret-safe
↓ pinned egress · api.hubapi.com
✅ output.id → {{steps.sync.id}}Mapped
SOAP, Databases & Files

The systems that were here before REST

Enterprise integration is mostly not JSON over HTTPS. FlowClick speaks to the rest through typed connections that keep credentials out of flows and put every call on the same audit trail.

  • SOAP 1.1 / 1.2 — structured header and body, XXE-safe XML codec, response mapping back to fields
  • JDBC databases — parameterized query and execute against your own database through a database connection: Postgres out of the box, with MySQL, SQL Server, and Oracle dialects when you supply the driver
  • Object stores and file servers — S3, Google Cloud Storage, Azure Blob, SFTP, FTP/FTPS, SMB/CIFS, and NFS through the filesystem step: read, write, list, verify, and apply WORM retention
  • Mail and SMS providers as connections — your SMTP, your Twilio number, or FlowClick's managed sender
  • Mappings — versioned, reusable structural transforms between any two shapes, with lineage recorded
soap → SAP PI: CreateSalesOrder1.2
database → warehouse (SQL Server)JDBC
filesystem → s3://exports/2026-09/WORM 7y
filesystem → sftp://bank/incomingSFTP
transform → mapping: edi850-to-orderLineage
Self-Building Integrations

No connector yet? Ask for one.

Describe the vendor API to a governed agent and it drafts a connector descriptor and fixtures as a proposal. The platform re-runs the fixtures itself, records a verdict, and waits for a person to approve. Only then does the connector go live — for your account, or published to the Marketplace.

  • Same loop for mappings (fixtures re-run) and flow fragments (synthesized scenarios re-run in a disposable copy)
  • Verdict recomputed at validate, approve, and promote — a caller can never assert a pass
  • Impact analysis and the authoring agent run attached for the reviewer
  • Pairs with the vendor emulator pattern: test against a fixture server before you ever have production credentials

The self-building loop →

 Proposal #812 — connector: freightco
🤖 Descriptor + 6 fixtures draftedProposed
↓ platform re-runs fixtures
🧪 6/6 passValidated
👤 Approved · promotedLive
Atoms — Hybrid Runtime

Run it on your premises. Govern it from the cloud.

An Atom is a lightweight FlowClick runtime installed behind your firewall and paired to an application. Flows you assign to it execute entirely on your network — against your databases, your file shares, your internal APIs — and only the result comes back.

The atom step

A subflow call that runs on a named Atom. Inputs go in, outputs come out; the flow definition, version, and audit stay in the cloud where you manage everything else.

The exec step

Run local command-line tooling on an Atom — build tools, git, CLI agents — through a named exec connection, with output captured into the flow.

Train models on-prem

Model training can be dispatched to an Atom so sensitive data never leaves the premises; the trained artefact registers centrally like any other version.

Enrolled, not exposed

Atoms enrol outbound with a credential you can reissue, keep a health view in the account, and are reached over an authenticated channel — no inbound ports on your network.

Assign and unassign flows

Decide per flow what runs where, from the UI or over MCP. Move a flow back to the cloud with one call when the constraint goes away.

Same governance

An Atom run is traced, audited, and effect-budgeted like any other. QA sessions capture its side effects instead of firing them.

Events In, Webhooks Out

Streams and callbacks, reliably

Consume Kafka, Amazon SQS, Google Pub/Sub, AMQP, polled endpoints, and reliable webhooks — one flow run per message, at least once, deduplicated, optionally ordered. Publish platform events to your systems as signed HTTP callbacks with retry policies.

  • Idempotency keys, retry budgets, dead-letter queues with replay, per-source ordering with poison handling
  • Inbound events can resume waiting durable processes by correlation key
  • Outbound event subscriptions with filters, secrets, custom headers, and delivery logs
  • Inbound webhooks with HMAC or Twilio signature verification and synchronous or templated responses

Event ingestion in depth →

 Ingestion Overview
kafka: orders.created · lag 0Healthy
sqs: returns-queue · 3 in DLQReplay
webhook: /hook/stripe · 1,204 todayVerified
outbound: erp.order.updated → 2 subscribersDelivered

Bring the system nobody wants to integrate

SOAP, an on-prem database, a vendor with no connector — we'll show you the path for each, including one that builds itself.

Start Free Trial Schedule a Demo