Kohral is the orbital control plane for always-on AI agents. Define, plot, and operate the whole fleet from one command deck — live conversations, fleet-wide telemetry, vaulted credentials, and rollouts that only happen when you say so.
Draft agents free in hosted mode.
Pay only when you roll one out.

An agent goes live in an afternoon. Six months later nobody can say how many are running, which credentials they hold, what they cost, or who approved the last change. The fleet becomes a rumour.
Kohral makes it legible. Modeled on orbital cartography: every agent is a plotted node, colour- and shape-coded by lifecycle, so a growing fleet stays accountable and under command.
Every agent plotted across its runtimes, coded by lifecycle. Running, drafting, degraded — readable at a glance instead of buried in a list.
Read what your agents are actually saying, as they say it. Step into any thread, follow the handoffs, and intervene when a reply needs a human.
Messages processed, health signals, and per-agent spend across the fleet — so the bill and the behaviour are never a surprise.
Keys are stored in the vault and referenced, never pasted into a prompt. Preflight verifies every key before an agent is allowed to leave the ground.
Nothing goes live by accident. Rollout stays blocked until intent is confirmed, capabilities are chosen, and every credential checks out.
A three-step interview gives each agent its role, persona, and runtime — so intent is written down once, not re-litigated in a prompt every week.
Every thread an agent holds is inspectable in real time, with the context it was given and the tools it reached for. When something goes sideways you see the sentence it happened in.


Fleet-wide throughput, health, and spend broken down per agent and per period. The signal you need to retire a quiet agent or defend a loud one.
Walk the guided anamnesis: role, persona, capabilities, runtime. The agent exists as a draft — free, grounded, and reviewable.
Reference vaulted keys instead of pasting them. Preflight verifies each one and names anything still missing before launch.
Confirm intent and plot the agent into the fleet. From that moment it's live, metered, and visible on the deck.
We run the control plane and the runtime. Draft as many agents as you like at no cost; metering starts the moment one is rolled out.
Run agents inside your own infrastructure and keep every conversation and credential on your side of the boundary. Kohral stays the deck; the runtime is yours.
Define, interview, and review as many agents as you want in hosted mode. Nothing is billed while an agent sits on the ground.
The command deck itself — fleet map, conversations, telemetry, and the credential vault for your whole team.
Metering begins at rollout and stops when you retire the agent. Per-agent cost is visible on the metrics deck the whole time.
Draft your first agent free, wire it to vaulted credentials, and roll it out when — and only when — you're ready.