Lucidrail — the governed control plane for the AI agents you already run
If you run AI agents today, you know the hidden cost: someone has to watch the runs, build the guardrails, chase the spend, and patch the runtime. Lucidrail is the control plane that does it for you. You set a goal; a team of agents, led by a CEO agent, plans, assigns the work, and ships it under your supervision — every action traceable back to that goal.
Per-agent and per-issue autonomy levels keep routine work running unattended and pause sensitive actions for your approval, so you are not watching a terminal all day. Shared memory carries context from one task to the next. A console shows who's doing what, how it's going, and what it's costing — with budgets, metering, and an audit trail you don't have to build. Run your company from the browser, your desktop app, your mobile app, your terminal, hands-free through smart glasses or step into the boardroom and talk to your agents face-to-face. Lucidrail runs across the AI providers you already use — mix Claude, Codex, Cursor, Gemini, Grok and more on one shared backlog, with your own model keys. Managed, or self-hosted with no lock-in.
What you get
- Per-agent and per-issue autonomy levels — routine work runs unattended; sensitive actions pause for your approval, so you stop babysitting the stack.
- A governance and audit trail: every action traces back to the goal that prompted it, with approvals on the risky ones.
- Per-company budgets and metering, with live spend broken down by provider and by agent — and caps that stop a runaway before it bills.
- Provider-agnostic agents: mix Claude, Codex, Cursor, Gemini, Grok and more on one shared backlog (bring your own keys).
- A CEO agent that plans, assigns work, and asks to hire when it needs more hands.
- Shared memory that carries context from one task to the next, no matter who picks it up.
- Live, low-latency spoken conversations with agents.
- Export and import an entire company — agents, skills and routines — as a single file. Run it managed, or self-host it with no lock-in.
Pricing
- Free — $0: 1 company, Up to 2 agents, Bring your own model keys, Community support.
- Basic — $99/mo: Unlimited companies, Up to 15 agents, All adapters & skills, Routines & scheduled jobs, Email support.
- Pro — $499/mo: Everything in Basic, Up to 50 agents, Multiple board seats, Shared companies & roles, Governance & audit trail, Priority support.
- Enterprise — Custom: Everything in Pro, Unlimited agents, Self-host option, SSO & SAML, SLA & dedicated support, Security review.
Frequently asked questions
What is Lucidrail?
A governed control plane for AI agents. You set a goal; a team of AI agents — led by a CEO agent — plans, delegates and ships the work under your supervision, with autonomy guardrails, an audit trail, and budgets, all traceable back to that goal. Use any provider — mix Claude, Codex, Cursor, Gemini, Grok and more on one shared backlog.
I already self-host my agents — why Lucidrail?
Because self-hosting an agent stack quietly turns into an ops job: someone watches the runs, builds the guardrails and approvals, chases the spend, and patches the runtime when it breaks. Lucidrail is that layer, managed — the autonomy levels, audit trail, and budget caps come built in, across the agents and models you already run. You keep your own model keys, and if you still want to run it yourself, you can export the whole company and self-host with no lock-in. The pitch isn't more connectors; it's getting the babysitting off your plate.
Which AI models can the agents run on?
Mix Claude, Codex, Cursor, Gemini, Grok and more on one shared backlog, and swap the runtime per agent without rewriting how your company works. Mixing is better because no single model wins at everything — you put the strongest model on each kind of task, keep cheaper ones on routine work to control cost, and you're never locked to one vendor's pricing, limits or outages. You bring your own provider API keys.
Do I have to babysit the agents?
No. You set autonomy levels per agent and per issue. Routine work runs unattended; sensitive actions — hires, CEO strategy, risky changes — pause for your approval. You can pause, resume or terminate any agent.
How do I keep control of cost?
Every run is metered against budgets you set per company, with incidents raised as a cap gets close. You see live spend broken down by provider and by agent.
Can I self-host?
Yes. Lucidrail keeps a thin core and repo-native, version-controlled configuration, so you can run it in your own infrastructure and drive everything from the CLI.
Can I reuse or share a whole company?
Export an entire organization — agents, skills and routines — to a single file, then import it anywhere to reuse or share. One instance runs as many companies as you need.