One key. Every model.
Total control.
The control tower for agentic AI. One team key for every provider: route each teammate and agent to the right model, share secrets safely, and cap spend live, all governed from Claude Code over MCP.
control surface
Everything your AI touches,
under one tower
AIRCTRL sits between your agents and every provider: one governed control surface built from three primitives.
Airspace
One key in. Every model out.
Airspace is the gateway every request flies through, the one place you control everything. It takes a single team key and clears it to the right model, by person, task, or agent, with live spend caps and instant failover, so switching provider never means touching code.
- One key for the whole team
- 100+ models behind one endpoint
- Zero rewrites to switch provider
Tower
One tower. Total control.
The Control Tower is the defense layer. It sets and enforces the rules: who can use which models, how much, and when, by person, role, and policy. Every call is cleared, audited, and attributed from a single place.
- Controls by person, role, and rule
- Live clearance, caps, and rate limits
- Every call audited and attributed
Secrets
Secrets that travel with the task.
Secrets is the vault. API keys, environment variables, database URLs, and OAuth tokens live encrypted in one place, scoped from team down to a single agent run. Values are injected at the edge at request time and never revealed to the agent, the repo, or your logs.
- One encrypted vault, AES-256
- Scoped team, project, task, agent
- Injected at the edge, never revealed
pre-flight checklist
From zero to cleared in under a minute
AIRCTRL runs where your agents already work, right in your terminal.
Install the CLI
One binary for your shell, your CI, and your agents. No SDK lock-in.
Connect over MCP
Register the MCP server and Claude Code discovers the tower and every tool it exposes.
One key, every model
Route every teammate and agent to the right model with per-key policies, scopes, and live spend caps.
control surface
Governed, from the first call
Keys, models, secrets, and spend, all cleared and attributed from one place.
Your agents hold the controls
Connect AIRCTRL to Claude Code over MCP. Agents read keys, switch models, and ship, without ever seeing a raw secret.
One key, every model, by policy
Route each person, task, or agent to the right model with per-key limits and live spend caps. Swap providers without touching code.
Secrets that travel with the task
Share environment variables and credentials across projects, tasks, and teammates. All encrypted, scoped, and revocable in one click.
Govern every aspect of your AI
Clear who can use which models, how much, and when. Every call is audited and attributed from a single tower.
charter · enterprise
Built for teams that move fast
Enterprise-grade performance, security, and compliance, without slowing your developers down.
Integrates with your stack
A drop-in OpenAI-compatible endpoint, an MCP server, and native SDKs. Keep the tools your team already uses.
Compliant by default
SSO and SAML, full audit trails, encryption at rest and in transit, and SOC 2-ready controls out of the box.
Deployed and supported
Run it self-hosted or in our cloud. Onboarding, migration, and 24/7 support handled with you.
The control tower for agentic AI is on final approach.
AIRCTRL is launching soon. Request early access and we will clear you for takeoff the moment the tower goes live.
An AIRCTRL product. Built on theo, a TheoVex company.