All work

Execute

An approval-gated agent for creating, inspecting, and operating deterministic business workflows.

Status
Live
My role
  • Agent product and interaction design
  • Proposal, approval, and execution architecture
  • Tenant isolation, safety controls, and operational visibility
Stack
Next.js, TypeScript, Supabase, Drizzle, Groq, Resend
Execute product landing page

The problem

Users need a faster way to create and operate automations without giving an AI agent unchecked access to workflows, integrations, or execution state.

Most automation agents make it difficult to know what the model inspected, what it intends to change, and whether a mutation can be executed safely for the correct workspace.

The approach

Built a bounded agent tool loop that inspects tenant-scoped state, converts requested mutations into expiring proposals, validates them, and executes only after explicit approval.

Turns natural-language requests into validated proposals, requires explicit approval before mutations, and reuses a strict workflow executor for observable actions.

Defining decision

Keep model reasoning separate from mutation authority: every write becomes a persisted, validated proposal that the user must approve before deterministic code executes it.
The tradeoff

Approval gates add one deliberate step to agent actions, but make mutations inspectable, idempotent, tenant-safe, and much easier to trust.

How it works

Step through the system. Each stage lists the safeguard that keeps it trustworthy.

Request

The user describes an operational intent through the agent interface.

Safeguard

Treats language as intent, not immediate authority to mutate workspace state.

1 of 5

Architecture notes

  • Bounded tool rounds, result-size limits, prompt-injection defenses, and provider fallback control the agent runtime.
  • Persisted proposals use validation, expiry, idempotent decisions, and stale-state protection before execution.
  • Failure findings are tenant-scoped, deduplicated by execution, classified deterministically, and stored with credential-redacted evidence.

Agent execution proof

Built and working

  • Tenant-scoped tools for workflows, executions, forms, contacts, and integrations
  • Persisted create, update, run, retry, cancel, and disconnect proposals with explicit approval
  • Failure monitor with deterministic classification, redacted evidence, and guided repair actions

Engineering signals

  • Internal-release feature controls
  • Per-call token and latency tracking
  • Atomic daily usage limits

What’s next

Use internal-release telemetry and failure findings to tighten tool selection, proposal quality, and guided recovery before general availability.