Governed work surface

Sidekick turns
policy into action

Use natural language to run governed workflows inside LumenFlow Cloud. Sidekick follows policy, routes risky actions through approvals, and leaves evidence behind.

Start freeExplore Cloud
Sidekick Chat
Review the pending release tasks and prepare approvals for anything that touches migrations.
I'll group the work by repo, flag schema changes that need approval, and collect evidence. Estimated cost: $0.42. Proceed?
Yes, go ahead.
Working... ✓ Release tasks grouped. ✓ Migration changes flagged. Evidence compiling.
Goal-driven governance

Set outcomes,
keep the controls

Ask for outcomes in plain language. Sidekick plans the work, estimates cost, confirms before anything risky executes, and tracks progress against your objectives — all within the governance model.

  • Policy-aware execution boundaries
  • Cost estimates before execution
  • Approval prompts for risky actions
  • Evidence captured automatically
Objectives
Release candidate ready for reviewRunning
75% · 4 tasks
Weekly evidence export deliveredComplete
100% · 3 tasks
Agent spend review closedRunning
30% · 5 tasks
Reusable workflows

Routines for repeatable governance

Schedule recurring tasks for the work that should always happen: nightly evidence bundles, weekly policy reviews, and release readiness sweeps that arrive with context already attached.

Nightly evidence export

Package the day's governed actions, approvals, and receipts before the next team handoff.

Weekly policy review

Summarize denials, exceptions, and approval load so platform teams can tune guardrails.

Release readiness sweep

Check migrations, test evidence, and unresolved approvals before a release moves forward.

Real-time

Reactive triggers
for risky events

Sidekick watches for the events that matter: failed gates, blocked secrets access, schema changes without migrations, and spend thresholds. When they fire, it responds inside your approved playbook.

Schema change without migration→ Hold release and request reviewer approval
Agent asks for restricted secret→ Block access, log denial, and alert workspace admin
Daily spend hits 80%→ Pause non-critical work and notify owners
Context & control

Memory that persists. Autonomy you control.

Sidekick remembers workspace conventions and operating context across sessions. Dial autonomy up or down per workspace without leaving the policy model behind.

Supervised

You approve everything

  • Sidekick plans and proposes
  • Nothing executes without your OK
  • Ideal for onboarding and high-risk work

Guided

Routine is automatic

  • Handles known patterns autonomously
  • Flags anomalies for your review
  • Most teams land here after a week

Autonomous

Full speed, full proof

  • Operates within budget and scope
  • Proof logged for every action
  • For mature workflows with guardrails
Guardrails & evidence

Enforced constraints.
Proof of every action.

Scope limits, budget controls, and approval workflows keep Sidekick inside the rules. Every action generates an execution receipt: what changed, why it was allowed, and who approved it.

Scope limits
Budget controls
Approval workflows
Proof of execution
What changedTask completed within governed workspace
Why allowedPolicy boundaries verified before execution
Who approvedApproval checkpoint passed for workspace-scoped action
What remainedExecution receipt exported to audit log
Flexible inference

Managed or bring your own

Start with LumenFlow’s managed inference — no API keys needed. When you’re ready, plug in your own keys for OpenAI, Anthropic, or any compatible provider.

Managed inference

Zero config. We handle model selection, scaling, and cost optimization. Pay per action, not per token.

Bring Your Own Key

Full control over model choice and spend. Use your existing enterprise agreements and volume discounts.

Bring Sidekick into
governed workflows

Start with one governed task, keep approvals where they belong, and expand into routines once the evidence model is working for your team.

Start freeExplore Cloud