✦ The stack

The applications that make Mira work.

Mira isn't a thin wrapper around one chatbot. She's an integrated system — a reasoning core, a rival auditor, a real memory, a live voice and video presence, creative engines, every major channel, and a measurement layer. Here's the actual application stack, grouped by what each part does for her. Best-in-class tools, wired together with judgement.

Reasoning & judgement

Two minds, not one — because self-grading is worthless.

Claude (Anthropic)

Mira's core reasoning brain — strategy, diagnosis, copy, and every decision in the loop. The latest, most capable models.

A rival model (OpenAI)

An independent auditor that adversarially reviews Mira's decisions and reports to you. Its only job is to find where she's wrong.

Memory & knowledge

So she compounds instead of starting from zero every session.

PostgreSQL

The system of record — every campaign, decision, lesson and timeline, isolated per tenant.

Jina embeddings (self-hosted)

Semantic memory — she recalls relevant past lessons by meaning, not keyword. Run on our own hardware for privacy.

Tavily

Live web research — market, competitor and trend signals fed into her context.

Voice & video presence

She's someone you can actually talk to — face to face.

Pipecat

The real-time voice pipeline — low-latency, barge-in-aware conversation, not walkie-talkie turns.

Tavus

Her live video avatar — the face on the "talk to Mira" call.

Sarvam & Cartesia

Natural text-to-speech, tuned for Indian and global English voices.

Twilio & Plivo

Her phone presence — inbound and outbound voice, plus WhatsApp.

Creative production

Creative is the main variable in modern marketing — so she makes it at scale.

Image generation

On-brand visuals generated to a creative brief, scored before they ship.

Creatomate

Templated short-form video and reels — the engine behind motion creative wiring in.

Channels & distribution

One brain, every channel a customer touches.

WhatsApp (Twilio)

Two-way conversation on the channel your customers actually open.

Brevo

Outbound email and lifecycle sequences at deliverable scale.

Gmail API

Mira's own inbox — she reads and replies on her own address.

Slack

Where she reports to you and you approve her moves with a tap.

LinkedIn & X

Organic posting, channel-native, via per-tenant connections you authorise.

Telegram

An extra direct line to Mira where you want it.

Paid & measurement

She spends against attribution she can see — never blind.

Meta Marketing API

Create, pause and read Facebook/Instagram ad campaigns inside your account.

Google Ads API

Search and display campaign control, with new spend gated behind your approval.

Google Analytics 4

Traffic, channels and conversion events — the raw signal behind her diagnosis.

Meta Pixel

Conversion tracking that ties spend back to outcomes per product.

Platform & security

Multi-tenant, encrypted, and built to keep your data yours.

FastAPI (Python)

The application core that wires the brain to every tool.

Redis

The fast queue and state layer behind her real-time work.

HashiCorp Vault

Per-tenant secrets and key encryption — your connected accounts are sealed, not shared.

Per-tenant OAuth brokers

You connect your own Google, Meta, LinkedIn and X accounts; tokens stay scoped to you.

Google Drive API

A shared document repository the chiefs can read and contribute to.

Automated TLS

Certificates renew themselves — encryption that never lapses.

Why list this at all? Because "an AI agent" can mean a weekend wrapper or a real system. This is the second kind. If you're evaluating seriously, the stack tells you whether the company behind the agent is, too.

Serious tools. One CMO.

See how it comes together — talk to Mira live.