CONFIDENTIAL — Prepared exclusively for PredictLab. Not for distribution.
×
PredictLab
LUNAR AI LAB

Get More Markets Live

Rewrite your agent's prompt chain, add new data sources, calibrate with real feedback.

Scope
2 Stages
Timeline
6–8 Weeks
Date
March 2026
WHO WE ARE

7 years in crypto.
300+ projects. Now building AI.

Lunar AI Lab is the build team inside Lunar Strategy, a European growth marketing agency with 7+ years in tech and crypto. 300+ projects delivered. $60M+ in managed marketing budgets.

We build custom AI agent systems for marketing and revenue teams. We also bring the marketing brain that makes agent outputs perform when they hit real users.

0
Years in crypto & Web3
0
Projects delivered
0
Managed budgets
Trusted by: Polkadot · OKX · Polygon · Cardano · DFINITY · BitMEX · Neo
Lunar AI Lab team at work
WHERE YOU ARE TODAY

Your agents work.
They need calibration to reach full output.

Current State

  • Governor rejects too many market ideas. The pipeline produces less than it should.
  • Google News is the only data source. No real-time signals from X or Twitter.
  • No handling for void markets, disputed outcomes, or early settlement
  • Nothing guides which markets are worth creating. The system has no "taste."
  • No human review step. The agents run on their own but nobody checks the output.
  • Creative and Publisher agents work fine, but they get weak input from upstream

What Changes

  • We rewrite the Sentinel → Architect → Governor prompt chain with better filtering criteria
  • X API added as a second data source for real-time trends alongside Google News
  • Edge cases handled: void markets, disputed outcomes, early settlement, volatility flags
  • Marketing knowledge added: niche market templates shaped by 7+ years of crypto marketing
  • Human review step added for calibration first, then we transition to autonomous with monitoring
  • Better input upstream means Creative and Publisher agents produce better output with zero changes to their code
THE SOLUTION

Better pipeline, better markets.
Autonomy with guardrails.

Data Sources

  • Google News (existing)
  • X API New
  • Crypto price feeds (existing)
  • Configurable RSS sources

Aura Agent · LLM Pipeline

  • Sentinel: news filtering & relevance scoring
  • Architect: market question generation
  • Governor: quality gate with calibrated criteria

Human Review Queue

  • Approval / reject / edit interface
  • Calibration data collection
  • Category tagging
  • Monitoring only Stage 2

Market Output

  • Creative agent generates images + copy
  • Publisher agent pushes to platform API
  • Telemetry logged to dashboard
DELIVERY ROADMAP

From kick-off to autonomous
in 6–8 weeks.

Week 0
Week 1
Week 2
Week 3
Week 4
Week 5
Week 6
Week 7–8
Kick-Off Workshop
Codebase Audit & Assessment
Stage 1 Development
Stage 1 Presentation
Deployment & Calibration
Stage 2 Development
Stage 2 Presentation
Stage 2 Deployment & Handoff
Meetings
Deliverables
Development
Week 0

Kick-Off Workshop

  • Alignment call with Enk (CTO) and Khish (BD)
  • Review roadmap, agree on market categories
  • Get codebase access, API credentials, internal docs
  • Define initial calibration criteria together
Weeks 1–3

Audit & Stage 1 Build

  • Full codebase audit, deliver written assessment
  • Rewrite Sentinel → Architect → Governor prompt chain
  • Integrate X API as data source
  • Build human approval queue + dashboard integration
  • Basic edge case handling
Weeks 4–5

Calibration Period

  • Stage 1 goes live with human review
  • Lunar + advisory team review every market before publishing
  • Collect calibration data: what gets approved, rejected, edited
  • Refine filtering criteria based on real performance
Weeks 6–8

Autonomous & Handoff

  • Bake calibration learnings into prompts
  • Add niche market templates + category-specific rules
  • Remove human approval gate → autonomous with monitoring/alerting
  • Documentation, handoff to Enk, team walkthrough
SCOPE OF WORK

What's included.
What's not.

What We're Building

  • Refactor Sentinel → Architect → Governor prompt chain
  • X API integration as real-time data source
  • Human approval queue + dashboard integration
  • Edge case handling (void, disputed, volatile markets)
  • Agent reliability testing (output quality + edge case coverage)
  • Niche market templates and category rules
  • Calibration support during human-review phase
  • Full documentation and handoff to Enk

Out of Scope

  • Platform security audits or penetration testing. We test agent reliability, not platform infrastructure.
  • Rebuilding Creative or Publisher agents (these work as-is, no changes needed)
  • Platform UX/UI changes or redesigns
  • Ongoing development beyond the retainer scope
  • Community building, marketing activities, ambassador programs, and content strategy. As agreed with Adam, these go into the advisory package as a separate workstream.
  • Any additions scoped and quoted separately

Note: Our testing scope covers agent reliability (false positive/negative rates, edge case handling, output quality validation). Platform-level security audits and penetration testing are outside this engagement.

INVESTMENT

One-time build. Optional retainer.

Fixed price for the full pipeline development. Monthly support available after handoff.

$5,000

Agent Pipeline Development

Fixed price. Stage 1 + Stage 2.

Full Aura agent development: prompt chain rewrite, X API integration, human approval queue, edge case handling, calibration period, autonomous transition, documentation, handoff.

$500/mo

Post-Handoff Maintenance & Advisory

Monthly retainer. Starts after Stage 2 handoff. Cancel anytime.

Up to 8 hours/month of ongoing agent support, prompt tuning, and optimization.

What the Retainer Covers (up to 8 hrs/month)

  • Prompt tuning and optimization based on market performance data
  • Adding new market categories or niche templates
  • Adding new data sources (RSS feeds, additional news APIs)
  • Adjusting Governor filtering criteria as market patterns evolve
  • Edge case handling updates for new market types
  • Monitoring review and alerting threshold adjustments
  • Priority Slack/Telegram support for agent-related issues
  • Monthly performance review call (30 min) with recommendations

New feature development, platform-side changes, and agent architecture rebuilds are outside the retainer and quoted separately. No lock-in. Cancel anytime with 30 days notice.

Payment Structure

$5,000 · Pipeline Development Invoiced upfront before kick-off workshop
$500/mo · Retainer Invoiced monthly, starting the month after completed handoff
PROOF OF WORK

What we've built. What it did.

ContentOS · AI Content Engine

3-agent pipeline (Researcher → Idea Maker → Writer) built for a crypto marketing team. Trained on existing brand content and voice guidelines. Pulls from X, Google News, and industry sources. Generates platform-ready posts with hooks and formatting.

  • Brand voice training on existing writing style, tone, and formatting
  • Multi-source research from X, Google News, Reddit, and niche industry feeds
  • Content hook generation matched against brand content pillars
  • Platform-specific formatting for X, LinkedIn, and blog
  • Engagement optimization based on post-performance feedback loops
400K+
impressions in 14 days
6.5%
engagement rate
3
autonomous agents
Multi-Agent Pipeline

RevenueOS · BD Meeting Intelligence

Full pipeline from calendar booking to same-day proposal. Pre-call research briefs delivered via Slack, automatic Fireflies transcript retrieval, post-call analysis, and proposal generation. Built for a BD team running 15+ calls per week.

  • Auto-triggered on calendar booking. No manual input needed.
  • Company intel, competitor research, funding data, and key people pulled automatically
  • Structured brief delivered to rep's Slack DM before the call
  • Fireflies transcript retrieval and post-call analysis with action items
  • Same-day proposal drafting from meeting context and research data
20%
reduction in BD admin
5 min
pre-call brief delivery
Same-Day
proposal turnaround
Workflow Automation

SignalOS · Competitive Intelligence

Automated competitor monitoring across web, Reddit, and Hacker News. Multi-layer signal pipeline: collection → filtering → classification → Slack delivery. Replaces two hours of manual feed scanning every morning.

  • Multi-source monitoring across 40+ configured feeds, updated every 30 minutes
  • Signal relevance scoring by type: launches, pricing changes, partnerships, hiring
  • Automated classification drops irrelevant noise before it reaches anyone
  • Morning Slack digest at 8:30 AM with top signals ranked by relevance
  • Trend detection and weekly competitor activity summaries for leadership
12+ hrs
saved per week
50+
signals captured weekly
3 min
morning digest delivery
Live & Running
NEXT STEPS

How to get started.

1

Review this proposal internally

Share with your team and Enk. We're available for any technical questions via the group chat.

2

Confirm scope and sign off

Once aligned, we lock in the timeline and send over the agreement.

3

Kick-off workshop

Joint session with Enk, Khish, and Lunar team. We align on roadmap, get codebase access, and define market categories.

4

We start building

Codebase audit in Week 1. Stage 1 live by Week 4. Full autonomous handoff by Week 8.

ABOUT LUNAR AI LAB

The build team inside Lunar Strategy. We build custom AI agent systems for revenue and marketing teams, then hand over full ownership.

Luka Mrkic · luka@lunarstrategy.com

lunarailab.com

This proposal is valid until 30 April 2026.