meerada

The AI operations company.
Measure what is real.

Every organisation now runs on AI models it did not build and cannot see inside. Meerada is the layer around them: we measure the models, operate them for you, protect your work while they run, verify what they say, and build the apps people actually use. Everything ships from one team, one codebase, one rule.
Open the exchange Get the app — free
Measure

📊 Exchange

Every model from every lab, graded hourly on tasks a program verifies, priced by cost per verified task. Open data.

The exchange →Grade board →
Operate

🎛️ Manager

One app for every model on your own keys. Import your history, run many sessions, switch models mid-conversation.

Manager →Handshake →
Protect

🛡 Guard

A watchdog on every AI call: stuck vs. working hard, token burn caught in seconds, a cage on what leaves the workspace. Company-wide policy.

Guard →
Verify

⚖️ Judge

Cross-check answers with juries from rival labs; a reliability score beside every answer; our own judge model in training.

Cross-check →Judge →
Build

🏙️ Tribes

The AI-powered city app: real people around you, real steps, real rooms, AI characters that speak your language — running on models we rank.

Tribes →
350+models tracked on the exchange
24 / 7hourly measurement, published as open data
5products in one download and one site
$0every product, open beta

Why one company, not five tools

The pieces feed each other. That is the design, not an accident.

The exchange feeds everything

The Manager's ⚡ Auto picks the top-graded model for your keys from the live exchange. Guard's "burning" threshold is the exchange's measured cost for that kind of task. Cross-check's jurors are chosen from labs the exchange knows. Tribes' characters run on models the exchange ranks.

Everything feeds the exchange back

Every Cross-check verdict is stored (in the user's tenant) and becomes training data for the Judge. Every Guard alert is a data point on how models fail in the wild. Every Handshake is a measured migration. The moat is not code — it is measurement history, neutrality and continuity.

Neutral by construction

We sell no tokens and take no lab's money for a grade. A model is never judged by its own family. Sponsored grading buys measurement, not results. The battery and the checkers are public; anyone can rerun them.

Honest by default

Small samples show wide intervals. A split jury says "unknown" instead of pretending. A thin panel is labelled as such. Grades that cannot be reproduced are not grades, so we do not publish them.

"Everything under the base model will be free. What holds value is what accumulates above it: measurement history, neutrality, a reliability ledger of every model, and the tools people work through every day."

What exists today

For partners, customers and anyone evaluating the company: the assets, in plain words.
AssetWhat it isStatus
The exchange (meerada.app)Hourly grader, public battery of verifiable tasks, CPAT pricing, Frontier Watch, labs & valuations, new-model feed, open JSON dataLive, hourly since Aug 2026
Manager (desktop + hosted)Multi-model cockpit on the user's keys; Claude Code / Claude.ai / ChatGPT import; the Handshake; fork, relay, attachments; ⚡ Auto routingv0.2.10 — Windows, macOS, Linux
GuardStall/burn detector (progress-anchored), cage on outbound text, per-user and company policy, Claude Code watcher, alerts to desktop / SlackShipped in the app and CLI
Cross-check + JuryNeutral panels from rival labs, frozen verdict schema, Fleiss' kappa, in-tenant verdict store = training setShipped; judge training planned
Handshake ReportMeasured migration of a company's real AI workload between models: where to move, what you save, where quality breaksEngine ready; delivered as a service
TribesAI-powered city app (Android live, iOS in review): radar, walking leagues with real steps, AI characters, TeebyLive on Google Play
Open-source engineVerification, metrics, battery, grader, CLI — Apache-2.0, 300+ testsgithub.com/ravemm-hub/meerada

Talk to us

Labs who want their model measured daily, companies who want a measured migration, teams who want Guard across the office, partners who see the whole picture.
Email Meerada   Read the code