A five-minute daily check
Six counts in one workspace say whether anything needs attention, and each opens onto the records behind it.
5 minutes
for the daily health check
Exactly once
posting, including on recovery
No payloads
in telemetry, and no secrets
The problem
Integrations rarely fail loudly. A message is dropped, a retry posts a journal twice, a file arrives malformed, and nobody knows until a customer calls or the books will not reconcile. Finding out what happened usually means a developer, a database query and an afternoon, and fixing it means re-keying data by hand and hoping nothing was posted twice.
How Lattice answers it
Lattice treats running an integration as part of the product. The integration operations workspace opens on six counts that tell you whether anything needs attention: dead letters awaiting action, failed service calls in the last day, inbound orders on hold, stale idempotency claims, claims needing manual reconciliation and unresolved business event errors. Each count opens straight onto the records behind it.
A failed message lands in a dead-letter queue, not the void. From the operations console you can retry it, hold it until an upstream cause is fixed, correct the payload and resubmit it, or discard it, while a scheduled sweep retries the rest on its own. Recovery reuses the original request id, so even a retry of something that half-succeeded cannot post twice. If a crash ever leaves a call stranded between creating and posting a document, a reconciliation sweep finds it and resolves it, or parks it for a person.
When someone asks what happened to the file a partner sent, the inbound message inquiry follows it from file to record to service call to dead letter, with the payload downloadable at every stage. And when you need to reproduce a problem, the service test console, the file ingestion simulator and the Liquid template tester let you do it without guessing.
Security follows least privilege: every operation has its own privilege, and the administrator, operator and read-only roles separate who configures integrations, who runs them and who only watches. The operator role leaves out high-risk operations such as reversals and cancellations. Every call is logged with its correlation ids, logging detail is tunable per operation, and telemetry records counts and durations, never payloads or secrets.
What you get
Found, fixed and proven, without a developer.
What is in the box
The operator's toolkit.
Everything here runs inside Finance & Operations, under its own security.
Reliability guarantees 5
- Idempotent operations keyed on the caller's request id
- Dry-run preview for ledger journals and sales order postings
- Self-healing sweep for calls stranded by a crash
- Automatic retry of failed asynchronous work
- Business events sent only after the transaction commits
Monitoring and recovery 6
- Integration operations workspace with six health counts
- File ingestion operations workspace
- Dead-letter console: retry, hold, release, fix and resubmit, discard
- Inbound message inquiry across files, records, calls and dead letters
- Service call log and statistics, tunable per operation
- Business event log with delivery errors
Testing tools 4
- Service test console with saved requests
- File ingestion simulator for a single payload
- Liquid template tester on the runtime engine
- Guided wizard that tests a file connection before saving it
Security 6
- A privilege per service operation
- Administrator, operator and read-only roles
- High-risk operations kept out of the operator role
- Every call logged with correlation ids
- File connection secrets held in Azure Key Vault
- Telemetry with counts and durations, never payloads or secrets
Deployment and lifecycle 4
- Azure API Management bundles: Bicep, APIOps or ARM
- Data management configuration captured to source control and reapplied safely
- Service metadata refreshed automatically when the server starts
- Each customization can be verified and turned off on its own
Questions
What buyers ask.
Who runs this day to day?
An integration administrator or operations analyst, not a developer. The daily check is read-only and takes about five minutes; recovery is a few clicks on the console.
What if Finance & Operations stops in the middle of a posting?
The idempotency claim is written in the same transaction as the document. A scheduled sweep finds any claim left in progress and resolves it: finalized if the document posted, released if it never existed, or parked for review if it was created but not posted.
Do our payloads end up in logs we don't control?
No. Telemetry carries counts and durations only, never payloads or secrets. Payloads stay in your environment, in the logs and queues you can see and secure.
Can we switch part of it off?
Yes. Each process customization can be verified and turned off on its own, and asynchronous operations can be disabled individually.
Try it
See it in your own sandbox.
Lattice is one package: every capability on these pages comes with it. Run it free in any non-production environment for as long as you like, and subscribe only when it goes to production.