Every model your company uses,
in one place you control
Put OpenAI, Anthropic, Google, Amazon Bedrock, Azure, and your own models behind one gateway that runs in your environment. Decide who uses what, see what it costs, and keep a record of every call.
Any model,
including your own
Use frontier models from public providers, models in your own cloud account over a private connection, or open-weight models on your own GPUs.
Public providers
OpenAI, Anthropic, Google, Amazon Bedrock, and Azure, behind one gateway.
Your cloud account
For example Amazon Bedrock over PrivateLink, with an IAM role instead of stored keys.
Your own GPUs
Open-weight models, tuned by xpander so your people get strong results from the models you host.
Change models without changing your apps or agents.
What you control
from one gateway
Decide who uses what, see what it costs, and keep a record of every call.
Model access
Choose which groups, teams, and agents can use which models. Groups can follow your identity provider’s groups.
Budgets
Set limits for each group, member, API key, agent, and even a single task, so a large analysis job never costs more than you allow.
Cost visibility
See spend by team, model, agent, and end user, including usage on licenses you bring yourself.
Keys
Provider keys stay in your vault. Apps, agents, and laptops never hold them.
Routing
Send each request to the right model: a public provider, a model in your own cloud account, or your own GPUs.
Logs
Every request is recorded with the user, agent, model, and cost.
Already have an AI gateway? Keep it
Keep the gateway you run today for routing, failover, and caching. xpander governs it alongside its own.
Route to vendors
Fail over between models
Retry failed calls
Rate-limit traffic
Balance load across keys
Cache responses
Log requests
Track usage
Serve your other apps
Run in your network
Scope model access
Use IdP groups
Cap agent spend
Budget each task
See cost by team
Vault your keys
Route to your GPUs
Log every call
Export the audit log
Connect coding tools
Runs in your cluster,
connects your coding tools
Deploy from a Helm chart into your cluster, with scaling built in: in your cloud account, on-premises, or air-gapped. Then connect each laptop’s coding tools with one copied command.
Helm chart
Deploy into your cluster, with scaling built in.
Anywhere you run
Your cloud account, on-premises, or air-gapped.
Coding tools in one step
Claude Code, Codex, and OpenCode connect with one copied command per tool. Your own code connects with a copied prompt.
No provider keys on laptops
Coding-tool calls get the same access rules, budgets, and logs as everything else. Provider keys stay in your vault.
Know where every AI dollar goes,
and cap it
Spend by agent, task, model, and person in one view, with budgets that warn or stop at the thresholds you set.
Questions we hear from IT leaders
OpenAI, Anthropic, Google, Amazon Bedrock, Azure, and self-hosted models, including open-weight models on your own GPUs.
The gateway runs in your environment. Requests go only to the providers you choose. With models in your cloud account or on your own GPUs, nothing leaves.
Per group, member, key, agent, task, and end user.
Same platform
What comes next
The rest of the platform
Add the MCP Gateway for tools, then the Agent Runtime and AI Workspace, on the same platform.
