Skip the workflow engine. Ship the case workflow.
Latch gives platform teams and startups the case lifecycle, governed approval workflows, and immutable audit logging out of the box. Write plugins in TypeScript or Go, bring your own AI models, and deploy on-prem or in the cloud. You build the parts that are unique to your business — not the plumbing.
Build on the platform, not around it
Startups building internal ops tooling
Teams that need approval workflows and audit trails from day one — without building them from scratch or stitching together five SaaS tools.
Platform engineers extending case workflows
Teams writing plugins to connect internal APIs, ERPs, payment processors, or custom ML models to a governed case lifecycle.
Teams bringing their own AI models
Teams that want intelligent triage and classification without handing case data to a third-party model provider. BYOM keeps inference inside your boundary.
Four ways to make Latch yours
The platform is designed to be extended, not forked. Plugins, models, and deployment are all first-class configuration points — not afterthoughts bolted onto a closed system.
Plugin SDK in TypeScript and Go
Write discover and execute functions in the language your team already uses. The SDK handles the contract with Latch — serialisation, auth, response formatting — so the plugin stays focused on the action.
Bring your own models
Plug in your own classification, routing, or risk-scoring model. Latch wraps it in the same governed workflow and immutable audit layer as everything else. Swap models without changing the workflow.
On-prem and private cloud deployment
Run the entire platform inside your own environment. Case data, models, and identity never leave your boundary. Start managed and self-host later if requirements change.
Governed plugin execution
Every plugin action inherits the platform's permission model, approval gates, and immutable logging. You write the business logic — Latch handles the controls.
What teams ship in the first week
These are not aspirational roadmap items. They are real workflows that teams have stood up in days using the SDK, a plugin, and the built-in case lifecycle.
Custom refund pipeline
Write a plugin that calls your payment processor, wrap it in a maker-checker approval gate, and get an immutable audit log on every execution — live in days, not quarters.
Internal fraud review workflow
Bring your own risk-scoring model for case triage. Latch routes flagged cases to the right reviewer, enforces segregation of duties, and records the decision trail.
Vendor exception handling
Build a plugin that updates your ERP after a governed approval. The case carries the full context — who requested, who approved, what changed, and what the downstream system returned.
Ship governed workflows in days, not quarters
The case lifecycle, approval engine, RBAC, and audit log are already built. You write the plugin that makes it yours.
Your models, your data, your environment
Deploy on-prem so case data and model inference stay inside your boundary. No forced dependency on external AI providers.
Every plugin inherits the full control layer
Permissions, approval gates, and immutable logging apply automatically. You never build a separate auth or audit system per integration.
Questions platform teams and startup engineers ask
These questions come up when teams are deciding whether to build their own workflow engine or extend an existing case management platform.
How is this different from building our own workflow engine?
You skip the hardest 80%. Latch gives you the case lifecycle, the approval engine, RBAC, maker-checker, the immutable audit log, and the plugin execution contract out of the box. You write the parts that are unique to your business — the plugin actions and the model logic.
What languages does the SDK support?
TypeScript and Go. Both SDKs handle discovery, execution, auth verification, and response formatting. Your plugin code focuses on the action logic.
Can I run my own models without sending data externally?
Yes. Deploy Latch on-prem or in your private cloud and point it at models running inside your own boundary. The governed workflow and audit layer work the same regardless of where the model lives.
How do plugins get permissions and audit logging?
Automatically. Every plugin action inherits RBAC, approval gates, and immutable logging from the platform. You do not build a separate auth or logging layer per plugin.
Can I start with the managed service and self-host later?
Yes. The deployment model is separate from the workflow definition. Move from managed to on-prem without rewriting plugins or reconfiguring case workflows.
How hard is it to ship the first plugin?
Most teams ship a working plugin in a day. The SDK includes a local test harness that simulates case context and role checks without needing a running Latch instance.
Bring one integration and see how it works inside the case
Pick the system your team relies on most — a payment processor, an ERP, an internal API, or a custom model — and walk through the plugin lifecycle from discovery to execution.