K
Kertas

Agent Execution Platform

Turn workflows and tribal knowledge into reusable agent skills. Run them on a governed runtime that improves with every execution.

How It Works

1

Upload Skills

Package your expertise as agent skills (SKILL.md). Define workflows, quality gates, and subagent prompts.

2

Run Agents

Deploy skills to workspaces. Agents execute through a ReAct loop — observe, think, act, evaluate.

3

Approve & Improve

Review every action before execution. Your edits become training signal. Skills compound over time.

Platform Features

ReAct Execution Engine

Queue-based agent loop with quality gates, subagent dispatch, and trace capture.

Approval Per Action

Every non-safe tool call pauses for your review. Approve, reject, or edit before execution.

Workspace Isolation

Organization > Workspace > Execution. Multi-tenant with configurable auto-approve policies.

Decision Trace Learning

Every cycle logged. Human edits captured. Skills improve from real usage data.

Formalized Function Calls

LLM proposes, you approve, system executes deterministically. 25 tools across 6 namespaces.

Cloudflare Edge Runtime

D1, R2, Vectorize, Queues. @cloudflare/shell for virtual filesystem. Millisecond cold starts.

Real Results

From our first production deployment

Case Study

Pinnacle Legal Partners — Client Onboarding

A 45-person Malaysian law firm needed standardized onboarding documentation for new clients. Previously took a senior associate 2-3 hours per client. Our agent produced a complete personalized package in under 5 minutes.

Welcome guide tailored to legal industry + Malaysian market
15-question FAQ covering PDPA compliance, Clio integration, Bar Council requirements
4-week onboarding checklist with role assignments
Handoff note flagging assumptions for human review
16
Agent cycles
4
Artifacts produced
29K
Tokens used
<5 min
Total time

Simple Pricing

Pay for what you use. No subscriptions required.

Agent Execution
1 credit
per 1,000 tokens
  • • ReAct loop cycles
  • • Subagent dispatch
  • • Tool execution
Most Popular
Starter Pack
10,000
credits for RM 400
  • • ~300 agent executions
  • • All tool namespaces
  • • Workspace isolation
Managed Service
Custom
skills distillation + operations
  • • Workflow audit
  • • Custom skill creation
  • • Ongoing optimization

Service First. Runtime Second. Platform Third.

We capture institutional know-how, operationalize it into governed agent skills, and continuously improve through decision-trace learning.

Get Started