Your clients. One workspace.
Set up a client, give them access, and help them run their first agent. Each client sees only their own work.
Sign in to manage OMNIA
Administration is for the person running OMNIA. If you were invited to a client workspace, open My agents instead.
Create your first client
Enter a name below. We will create their workspace and access key together.
How do I give someone access?
After you create a client, copy the key shown once. Send it privately with the Start here link. They sign in, open My agents, and see only their workspace. Adding an email does not send an invitation automatically.
Open the sign-in pageAdd an agent
What the scheduled runs use
Edit the spec (JSON)
Recent runs
| Started | Trigger | Status | Cost |
|---|
Client secrets (vault)
The client's own API keys for their tools. Agents use them by name ({{secrets.NAME}} in an http step's headers, or auth). Values are encrypted and never shown again.
| Name | Fingerprint | Updated |
|---|
Connect from Claude or Cursor (MCP)
Every active agent in this client's account appears as a tool. Use a key with Everything or Agents only access; runs count against the same budgets.
Replace CLIENT_KEY with the client's key. In the Claude API, pass the URL as an MCP server with the key as its authorization token.
| Label | Access | State | Last used | Expires |
|---|
Plan
Monthly statement
Learning dataset
Off by default. Scores, costs and models are always recorded under an anonymous ID; text is kept only when this is on. Their runs also appear as Wild Card rating cards.
| Agent | Runs | Spend | Agent budget |
|---|
| When | Action | By | Target |
|---|
Hyperagent
What this client's Hyperagent agents are doing.Advanced · System health, learning data & specialist tools
System health
Recent alerts
Wild Card
What OMNIA's self-improvement agent pushed for you. Every answer goes into the OMNIA Learning Dataset.Smell
Electronic-nose samples flowing into the OMNIA Learning Dataset.Devices
Recent samples
Fingerprint
Connect a sensor
- Hardware: a Bosch BME688 board (Pimoroni or Adafruit breakout, or the Bosch dev kit) on a Raspberry Pi over I2C.
- In Clients, add a client (e.g. "OMNIA Sensors"), open Keys and issue a key with access Learning dataset (write-only, for sensors). It can only send records.
- On the Pi: copy
scripts/odor-sensor/, thenpip3 install -r requirements.txt. - Save the key in a file only the Pi user can read:
install -m 600 /dev/stdin ~/.omnia-dataset-key(paste, then Ctrl-D). - Run:
- Samples appear here within a minute. Try it first without hardware by adding
--simulate.
OMNIA Learning Dataset
API data performance
Timings appear as data flows through the API.
JSON batches · streaming records · verified source retrieval. Connect your service →
Prove the difference.
Define success before collecting results. Compare the same cases, keep the sources, and see where OMNIA helps—or falls short.
Lock your evaluation
Compare the same case
Save or open a plan to begin.
Use the same inputs and conditions. Keep quality, time and cost as separate pilots. Have people rate actual agent outputs in the client portal; this form does not create human labels.
Open client portal →Results with their sources
A target is assessed only when all registered cases have valid pairs. This is a descriptive comparison, not proof of causation or a release approval.
Advanced: compare unregistered measurements
Compare the same cases under the current process and OMNIA. Missing or ambiguous pairs are shown; test data is excluded.
How to report pilot measurementsKnown issues (from the datasheet)
Claim IntelligenceTrace the statement. Inspect the sources. Follow the outcome.
See what connects.
One continuous path from a public statement to reusable knowledge.
Give your data a track record.
Track a claim to see real sources and outcomes connected here. No example findings are mixed into your data.
Owner-reviewed outcomes · Connections show recorded lineage, not causation. Motion is decorative; it does not represent measured EFVT or live transfers.