SYS · TALOUS · v3 EDGEPOINT · MYT
LOADING COMMAND
00%
CALIBRATING GUARD PRESS TO ENTER
TALOUS
Request demo
SYSTEM ONLINE
04 BRANDS 1,247 ADS
cycle 4 / 4 · 21:00
SCROLL TO ENGAGE ↓
Talous
Enterprise intelligence for ad operations

Command your ad spend.

The AI commander for ad agencies. Audits Meta on a cron, predicts outcomes, recommends action — with operator approval and a graded track record.

Request a demo Witness the loop
01 · Surfaces

Two halves. One commander.

VERDICT · PAUSE ad_88421
PRED · 7d 1.2× → 0.4×
CPL $58.20 FREQ 4.3 CONF 0.87
01

Manage the ads.

Sync Meta. Claude Guard predicts and verdicts. Operators approve from web or Telegram. Talous executes — then grades itself days later.

  • Audit cycles every 6 hours · tier classifier · fairness gates
  • Two-stage AI cascade — Haiku diagnostician, Sonnet decider
  • Outcome snapshots 1d / 3d / 7d / 14d / 30d
  • Brand memory consolidates weekly
JARVIS ~30 TOOLS · SONNET 4.6
how's Aurelia spending this week
$4,310 across 12 active ads · ROAS 3.4× · 1 pending verdict (BOOST ad_91120 · 38% headroom). Want me to approve?
approve it
✓ Approved · executed against Meta · grading queued for 7d
02

Run the agency.

The other half is the agency's nervous system — projects, schedules, CRM, journals. Jarvis sits across it as a thirty-tool AI command line.

  • Projects, tasks, workflows, AI briefs
  • Jarvis · web + Telegram · personas switch tool surface
  • Time-block scheduler with Google Calendar sync
  • Lightweight CRM · invoices · payments · Drive folders
02 · The Loop

Five movements. One discipline.

Every verdict feeds the next.

01

Sync

Meta Insights pulled per brand · 4× daily.

02

Guard

Claude reads · predicts · verdicts.

03

Approve

One tap · web or Telegram.

04

Execute

Action runs · journal stamps executedAt.

05

Grade

1d / 3d / 7d / 14d / 30d outcomes written.

03 · Track Record

Every prediction graded.
Every miss remembered.

Most AI tools forget. Talous keeps score. Each verdict becomes a row. Each window that passes becomes an outcome. The system that judged the ad now judges itself.

auto-generated brand narrative · re-injected into next Guard prompt
KILL82%
12 executed · median delta −0.18
PAUSE74%
19 executed · median delta −0.21
BOOST38%
8 executed · median delta −0.45
DUPLICATE61%
14 executed · median delta −0.09
last 90 days · per-action window talous · guard · v3
04 · Capabilities

Built for the room.

Every surface Talous ships is an operator surface. Nothing is decorative.

01

Claude Guard

The AI commander. Two-stage cascade with evidence gate.

02

Decision Journal

Every verdict — approved or rejected — journaled with reasoning.

03

Threshold Suggestions

The system watches its rejections. Suggests new thresholds quietly.

04

Telegram approvals

Inline-button verdict cards in your team chat. Audit-logged.

05

Jarvis · Work AI

Sonnet 4.6 with thirty tools. Personas switch the tool surface.

06

Self-hosted

One Docker container per agency. Your data, your servers. SL-1 → SL-17.

05 · For Agencies

Sized for the room you actually work in.

Talous is built for small ad agencies and in-house teams. Five to fifty brands. Three to fifteen operators.

Sized
5–50 brands per install · 3–15 operators
Owned
Self-hosted container — your hardware, your data
Scoped
Personal + team workspaces · roles · MFA
Wired
Meta today · Google Ads and TikTok stubbed
WAR ROOM · TODAY · AURELIA CYCLE 4 / 4
  • PAUSEad_88421CPL 81% over · freq 4.3approved 12:04
  • BOOSTad_91120ROAS 6.2× · headroom 38%auto · tier A
  • KILLad_7701814d ROAS 0.3× · evidence ✓approved 18:21
  • DUPL.ad_94455winner · CPL $11 · CTR 3.4%pending
  • PAUSEad_61203creative fatigue · cascadeapproved 14:50
OUTCOME GRADING · 1d / 3d / 7d talous · v3
06 · Take Command

Take command.

Demos run live against a real Meta account. Bring your worst-performing brand — we'll show you the verdict in fifteen minutes.

Request a demo

or write to [email protected]