Governance and security
Decide who reaches what, prove what happened, and put limits around AI before it reaches your customers.
Members, invitations, workspace roles and permission scopes, plus single sign-on through your own identity provider.
Keep a connection private, share it with a workspace or grant it to named people. Restrict which components workflows may use at all.
A record of who changed what, when and where, with a before and after for every change and one correlation id per bulk operation.
Keep credential material in HashiCorp Vault or AWS Secrets Manager instead of in ByteChef’s database. Workflows resolve secrets at execution time; nothing changes in the workflow itself.
One workspace-level content policy applied to every LLM call the instance makes, rather than guardrails configured per agent: redaction, blocked terms, moderation and injection detection, on the way out and on the way back.
Scale, reliability and observability
Grow from one node to a cluster without changing platform, and see what every run did - in ByteChef, or in the observability stack you already operate.
Run the platform as independently deployable, independently scalable services rather than one process.
Full OpenTelemetry observability: traces, logs and metrics over OTLP, from API calls down to individual component invocations, correlated under one trace id per run. No ByteChef-specific agent - if your stack speaks OpenTelemetry, you point ByteChef at it.
A rule fires when a run reaches a given status. Email delivery is in every edition; Enterprise adds webhook and Slack targets, so a failure reaches the system that owns your on-call rota.
Ephemeral, database-free workflow execution for bursty or isolated workloads.
Extensibility
Go past the built-in catalog, in the language your team already writes.
Author and publish your own components, or turn an OpenAPI specification straight into one.
Define workflows as code in Java, JavaScript, Python or Ruby.
Activate chat and embedding models once per environment, with credentials managed centrally rather than per workflow.
Collaboration and DevOps
Let more than one team build at once, without one team reaching another team’s production.
Group projects and control who reaches them, with roles applied per environment.
Connect a workspace to a Git repository. Projects push their workflows on publish and can be pulled back from a branch, so a workflow change reviews like any other change.
Development, Staging and Production, each with its own connections, credentials and configuration.
Embedded iPaaS
Offer your own customers integrations that run on your infrastructure, under your brand.
A connect dialog, an embeddable workflow builder, per-user credentials and per-tenant isolation, shipped inside your own product.
