Council · Infrastructure · MODULE COUNCIL
ONSET Council — Autonomous Goal Agent
Set a goal. Council figures out which AI Employees to call. Done.
Most AI tools answer single prompts. ONSET Council is goal-directed: tell it "recover this churning customer" or "onboard this new client". It plans the work, calls the right AI Employees (Module N, Module W, Module X, etc.), tracks progress, and reports back. Layer 4 of the ONSET stack — sits above every other module.
Goal-directed
Plans, executes, self-corrects
Cross-module
Calls every other Employee as needed
Audit-trailed
Every action hash-chained
WHAT THIS FIXES
The exact problems ONSET Council — Autonomous Goal Agent is built around.
- You have 25 AI Employees, each excellent at one thing — but no one orchestrates them as a team.
- "Recover this customer" requires Module N proposal + Module W reminder + Module A WhatsApp + manual coordination across all three.
- Single-prompt LLMs (ChatGPT, Claude desktop) can't plan multi-step work across systems and remember state.
- You spend 30% of your time on coordination work that could be automated.
HOW IT WORKS
The pipeline, end to end.
- 1
Set a goal
Plain English: "Onboard this new client" or "Recover the churning customer in Module W at 60 days". Council parses intent + constraints.
- 2
Plan generation
Council generates a 3-7 step plan with dependencies, expected outputs, and acceptable cost ceiling. Telegram preview with 30s cancel.
- 3
Execution
Each step calls the appropriate AI Employee (Module A / N / W / X etc.). Council waits for completion, checks output, decides next step.
- 4
Self-correction
If a step fails (LLM rate limit, retry needed, data missing), Council retries or adapts the plan. Within budget guardrails.
- 5
Report
Goal complete or stuck → Telegram report with what was done, what worked, what didn't, suggested next action by a human.
WHAT'S INCLUDED
Every capability shipped, named.
- Autonomous goal-directed planning + execution
- Calls every other ONSET AI Employee as needed
- Per-goal budget ceiling (LLM spend cap, time cap, action cap)
- Hash-chained audit log of every action taken
- 30-second human cancel window before high-stakes actions
- Self-correcting on transient failures
- Class A change management — major plan deviations require human approval
- Goal templates (onboard / recover / audit / launch / migrate)
- Cross-tenant for agencies (run goals for multiple clients)
- Telegram interface — set goal, get reports, intervene anytime
PRICING
Plans that scale with use, not seats.
All prices in MYR, exclusive of SST. 14-day refund on monthly, pro-rated refund on annual. PDPA-aligned. Cancel anytime.
MOST POPULAR
Team
RM 9,999/mo
50 goals/mo · all modules · custom goal templates · 5 users
Talk to sales →WHAT YOU RECEIVE
The artefacts that land in your hands.
Plan + Telegram preview
Before execution: 3-7-step plan + dependencies + budget + 30s cancel window.
Action audit log
Every step Council took, with timestamp + module + cost + result. Hash-chained for tamper evidence.
Goal report
End-of-goal Telegram summary: what was completed, what blocked, suggested human follow-ups, lessons for future similar goals.
FREQUENTLY ASKED
The questions buyers ask first.
How is this different from ChatGPT with tools?
ChatGPT runs a single conversation; tools are stateless. Council runs goal-state machines: plan → execute → check → adapt → repeat, persisting state across days/weeks. Built for goals, not chats.Can it spend my money without permission?
No — Mandate 2.4. Every high-stakes action (sending a message, charging a card, publishing publicly) gates through a 30-second human cancel. Mandate 2.5: 60% margin floor on bundle pricing (Module N) enforced.What if it picks the wrong plan?
Telegram preview before execution. You see the plan and cancel if it's wrong. If you let it run and it goes off course, the budget ceiling (LLM cost + action count) stops it before damage.Goal templates — what kinds?
Built-in: onboard client, recover churn, audit module, launch new product, migrate from competitor. Custom: define your own goal template in Team+ tier.How does it not loop forever?
Each goal has a max-action budget (default 50 steps) + max-spend (default RM 50). Council halts at either limit and reports.Compliance with PDPA / Class A change management?
Class A changes (modifying tenant data, sending outbound messages, publishing) require human approval per CLAUDE.md. Council respects this and surfaces approval requests in Telegram.
Try ONSET Council — Autonomous Goal Agent on your real business.
14-day trial. No credit card. PDPA-aligned. Built for Malaysian SMEs by a Malaysian team in KL. Reginald and ONSET team reply within one business hour, Mon–Fri 9am–6pm MYT.
- PDPA-aligned
- Cancel anytime
- 14-day refund
- 1 business hour reply
SEE OTHER AI EMPLOYEES
The other 24 modules in the platform.
Each AI Employee is independently usable, but they compose. Run a demo and we'll show you which combination fits your business.