Terminal-first · Private by default · In your control
The AI coding pilot that lives in your terminal
kpilot pairs a terminal-first coding agent with a dashboard your team controls — sign-in, devices, usage, and audit history in one place. Your code stays on your machine.
npm i -g @tagrise/kpilot
kpilot — ~/projects/api
$ npm i -g @tagrise/kpilot@latest
$ kpi login
Open https://konsolepilot.ai/device
Code: KP7F-92QX
✓ Signed in as you@company.com
$ kpi --mode write "harden the auth flow"
◆ Explorer and security agents inspecting your repo…
◆ Permission requested: modify 3 files
✓ Tests passed · usage synced to your dashboard
Everything your team needs to run AI coding safely
Private by defaultScoped API keysDevice approvalsUsage you can seeOne dashboardTeam policyAudit historyPrepaid creditSign-in from the browser
How it works
Up and running in three steps
No infrastructure to stand up, no code to migrate. kpilot meets you in the terminal you already use.
Step 1
Install the CLI
One command installs the kpilot coding agent. It runs where you already work — your shell, your repository, your machine.
Step 2
Sign in, take the controls
Approve the device in the browser. Issue scoped keys for your team and decide who can do what.
Step 3
Start shipping
Ask for the change. kpilot plans, edits, and runs your tests behind permission gates — then records what happened.
Features
Built for the way you already work
A coding agent that lives in your terminal, works with your tools, and keeps you in the loop.
Private by default
Your files and terminal work stay on your machine. Requests carry only what is needed, and prompt or source text is not retained.
Spend you can see
Prepaid credit means you only pay for what you use, and you can see exactly what every session cost.
Permission-gated writes
Every file edit and command is yours to approve. Run read-only, ask-before-write, or auto-accept when you trust the task.
Team access, centrally run
Invite people, approve their devices, and issue scoped keys — without handing out a shared credential or turning into a web IDE.
Audit you can hand over
An append-only, tamper-evident log of operational events, with prompt digests instead of prompt text, exportable for review.
The right model for the job
Complex work escalates, simple work stays cheap. kpilot picks the best fit automatically — you never juggle keys or endpoints.
Security & trust
Your code and your credentials stay yours
The safeguards are built in — so you can move fast without giving up control.
Nothing retained by default
kpilot stores usage metadata, not your prompt text, source, tool arguments, or model responses.
Tamper-evident audit log
Every operational event is chained by hash. Modification or reordering is detectable; prompts are stored as digests only.
Sandboxed execution
Commands can run in a sandbox that keeps the agent within your project and its environment.
Policy that only tightens
Central team policy can restrict models and permissions, and it can never loosen a local decision.
Credentials stay out of reach
Sign in from the browser and issue scoped keys. Secrets never live in your repository config or your prompt.
You keep the final say
Read-only plan mode, approval prompts, and one-key stop make the agent an assistant — never an autopilot you cannot stop.
Pricing
Pay only for what you use.
No subscription. Top up credit and pay only for the work you do.
No subscription
Pay as you go
Prepaid credit
Create an account, top up credit, and run your first task. You only pay for the work you do, and you can see the cost of every run on your usage dashboard.
Create your accountWhat's included
- Terminal-first coding agent
- Dashboard for devices, keys, and usage
- Prepaid credit — pay only for what you use
- See the cost of every run
- Tamper-evident audit history
- Team access and central policy
- Encrypted, transport-neutral state sync
FAQ
Questions teams ask before adopting
Still unsure about something? Reach out and we will walk you through it.
How do I get started?
Create an account, install the CLI with a single command, and approve the device from your browser. Top up credit and you can run your first task in the same session.
Does kpilot send my code somewhere?
Your files and terminal work stay on your machine. Only the request needed for a task leaves the machine, and prompt or source text is not retained. Local usage metadata is kept so you can see what happened.
How is my usage billed?
Pay as you go. You top up credit and kpilot spends it at the best available rate for the task — you only pay for the work you do.
What platforms and toolchains are supported?
kpilot runs on Linux, macOS, and Windows with a modern Node.js toolchain. It works in your existing shell and repository — Git, ripgrep, and optional sandboxing are picked up automatically.
Can my team standardise on it?
Yes. Invite people, approve their devices, issue scoped keys, and apply central policy across the team. Every action still runs through the same permission gates.
Can I run kpilot on my own infrastructure?
The CLI runs on your machine, while model requests and account features go through the kpilot platform. kpilot is offered as a hosted service rather than something you deploy yourself.
Ready when you are
Bring kpilot to your team today
Set up in minutes and pay only for what you use, or talk to us about rolling kpilot out across your organisation.
npm i -g @tagrise/kpilot