Drop gents. The more you drop, the smarter they get.
Each agent lives on a resource. Drop more, they discover each other and form a mesh. Ask one question — the investigation traces causality across your entire infrastructure.
For SRE teams and platform engineers tired of being the integration layer.
Your tools don't talk to each other.
The API is slow. K8s shows nothing. Postgres shows nothing. Dashboards say green across the board.
"error: no correlation"Each tool sees one resource. Nobody sees the chain. You're the integration layer.
You open five tabs, SSH into three boxes, and spend 45 minutes finding what changed.
"time_to_root_cause: 45m"The investigation is manual. The tools are passive. You do the reasoning.
Your on-call rotation wakes up at 3am to stare at the same dashboards you built to prevent this.
"alert: page │ action: look"Alerts detect. They don't investigate. The human is still the runtime.
Drop. Mesh. Ask.
One question triggers agents across K8s, Postgres, VMs, and gateways. The mesh follows the thread.
Root cause chains with evidence. Not "something changed" — "deploy at 02:00 removed POOL_SIZE config."
Agents recommend specific commands. You say yes or no. They never act without you.
Resolved in 47 seconds.
Four commands. That's the setup.
$ ssh prod-cluster && pip install tp-ops
# 2. Login
$ tp-ops login --token tp_live_xxxxxxxxxxxx
# 3. Drop an agent (auto-detects resource type)
$ tp-ops drop
# 4. From anywhere, talk to it
$ tp-ops ask prod-cluster
Compute · Databases · Networking · Messaging · Storage · DevOps. 200+ agents. One CLI. Auto-detects on drop.
Six categories. Every resource covered.
Drop an agent onto any resource. It auto-detects its environment and specializes immediately. 200+ agents across six categories.
| Agents deployed | Mesh capability |
|---|---|
| 1 | Expert on one resource. |
| 5–10 | Correlated insights across a service boundary. |
| 20+ | Full infrastructure mesh. Any question traces through everything. |
| 200+ | Complete coverage. Every resource type, every cloud, on-prem. |
Two AI models. Both run on your hardware.
Ternary Neural Network
- 2,888 parameters · <1KB
- <1ms inference · integer math only
- Weights: {-1, 0, +1}
- Always on. Watches metrics, logs, connections.
- Learns this resource's normal patterns.
- Hot-swap weight updates. Zero downtime.
TernaryPhysics-7B
- 7 billion parameters · 4-bit quantized
- ~15 tok/s on commodity CPU
- No GPU required
- Powers conversation. Reasons about problems.
- Reads logs, metrics, configs, events.
- Builds root cause chains. Suggests commands.
Reads everything. Writes nothing without you.
Agents investigate autonomously. They never write, modify, delete, restart, or scale without your explicit yes. Every time. No exceptions.
Both AI models execute on your resource. Data stays on your infrastructure. No cloud dependency. No data in transit to third-party servers.
Uses your existing kubeconfig, DB credentials, and cloud tokens. Never copies or caches them. No secret store to manage.
Every conversation logged. Every action tracked. Export to your SIEM. Full forensic chain from question to execution.