Put AI to work on sensitive data, at the lowest cost your rules allow.
Wayreeve is the control plane for enterprises that want governance and control over their AI workloads. It sends every request to the most cost-effective model your policies permit, blocks the ones the policies prohibit, and keeps a record your auditors can trust. It runs in your environment, on the provider contracts you already have.
One checkpoint between your teams and every AI provider
Every request from every app passes through Wayreeve, inside your environment. It checks the rules, picks the best-value provider that's allowed, and records the decision.
Bring order to AI adoption chaos
Every new AI use case needs the same three answers. Without a shared control point, each team gets them by hand, project by project.
Developers want models now
Teams wire providers into their apps one at a time. Each one brings its own keys, limits and review queue.
Security needs to know where data goes
Which provider saw patient or customer data, under which terms? Answering means chasing logs across teams and vendors.
Finance watches the bill climb
Cheaper batch and off-peak capacity exists, but each provider sells it differently, and its data terms often differ too. So most teams pay full price.
One control point, three outcomes
Approve once, enforce everywhere
Security writes the rules once, such as "patient data only goes to providers with a BAA and zero retention". Wayreeve applies them to every request from every app, and stops anything that doesn't qualify before it leaves your network.
Spend less on every workload
Work that can wait goes to discounted batch and flex capacity, often around half of list price. Wayreeve picks the cheapest option that meets the deadline and the data rules, across all your providers.
Answer the auditor in minutes
Every request carries a record of the rule applied, where it went and why. Export the evidence for a review instead of reconstructing it.
More than a gateway
An AI gateway answers how a request gets through. Wayreeve also answers whether it should, where it may go, what it cost, and how you prove it.
Every gateway answers
- Which model serves this, and what if it is down?
- Who may call it, and how many tokens?
- Was the prompt or the response unsafe?
- Which tools may this agent use, and how often?
- Who is this user, and what is their role?
- What happened, in the logs?
Wayreeve also answers
- May this data leave at all, and under which contract? Provider terms are checked against the data class on every request, tool call and agent task. Unknown terms fail closed.
- Will the answer hold up in an audit a year from now? A tamper-evident decision record with signed checkpoints, streamed to your SIEM.
- Was this the cheapest capacity the deadline allowed? A planner across realtime, batch and flex; budgets count models, tools and sandboxes.
- Whom is this agent acting as, and inside what? Agents act as the signed-in user, in governed sandboxes, with approvals where you require them.
- Do the same rules hold for models, tools and agents? One policy, one record, one set of budgets across all three.
One set of rules for prompts, tool calls and agent tasks
| Model requests | Tool calls | Agent tasks | |
|---|---|---|---|
| Where data may go | Provider terms checked against the data class; the cheapest permitted route wins. | Each tool server's terms checked against the data class before the call. | Each upstream agent's terms checked; cards filtered to what the caller may see. |
| Who may act | Keys and roles per team, project and model. | Per-tool permissions; agents act only within the signed-in user's rights. | Grants per agent and per skill. |
| What is checked | Prompts and responses, including prompt injection. | Arguments and results; tool definitions pinned and scanned for poisoning. | Messages in both directions. |
| What you can prove | A decision record per request. | A decision record per call, approvals included. | A decision record per task. |
Budgets count all three. The same tamper-evident record carries all three to your SIEM.
Live in days, not quarters
No rewrite and no new provider contracts. Your teams keep the tools they use today.
- Connect the providers you already useAWS Bedrock, Azure OpenAI, Google Vertex AI, OpenAI, Anthropic, or models on your own GPUs.
- Set your rules in plain termsWhich kinds of data may go where, which teams may use which models, and how much each team may spend.
- Point your apps at WayreeveA one-line change. Existing OpenAI and Anthropic code keeps working.
Your data stays yours
- Runs in your cloud account or data centre
- Uses your provider contracts and keys; providers bill you directly
- Prompts, outputs and files never reach Wayreeve
- Keeps serving requests even if our service is unreachable
- Supports fully disconnected sites
Built for the teams who sign off on AI
Platform teams
- One endpoint for every model and provider
- Keys, limits and budgets per team and project
- Automatic failover between providers
Security and compliance
- Data rules enforced on every request
- Redaction and guardrails before data leaves
- Single sign-on, roles and a tamper-evident audit trail
Finance and FinOps
- Spend by team, project and model
- Budgets with alerts and hard caps
- Savings from cheaper capacity, reported
See it work
Take the tour
See how a team sets up providers, data rules, guardrails and limits, then follow five requests through Wayreeve: one routed, one redacted, two blocked by policy, and one over its limit.
Send your own requests
Ask us for a demo code and call a live Wayreeve endpoint from your own code, using the OpenAI or Anthropic SDK.
Common questions
Does Wayreeve see our data?
No. Wayreeve runs inside your environment. Wayreeve Cloud manages licences and keeps the provider terms current; it receives only health information and usage totals, never prompts or outputs.
Do we have to change our applications?
Only the address they call. Code written for the OpenAI or Anthropic SDKs works as it is.
Will this make us compliant?
No single product does that. Your compliance programme and provider contracts set the rules; Wayreeve makes sure every request follows them and gives you the evidence that it did.
Do we buy inference from you?
No. You keep your existing provider contracts and pricing. Wayreeve helps you use them better.
Can our DevOps teams drive it from Claude Code or Cursor?
Next. A scoped, audited management MCP server: tokens that expire and never exceed their owner, every change previewed before it applies, every action in the audit trail. Today, the console, CLI and API.
Start with a two-week pilot
Your environment, your providers, your rules. We agree the success criteria with you up front.