AI Gateway
One governed front door for all your company's AI traffic: model keys, budgets and logs per team, plus an approved MCP catalog with central credentials and an audit log.
Free 30-minute discovery call. We'll see whether this fits your team.
$1,200 USD per install.
Every team calls models and MCP servers its own way: personal API keys, tokens pasted into config files, no idea who spends what. The AI Gateway puts one governed front door in front of all of it, for models and for tools.
On the model side, each team gets its own keys, budget and allowed models, and every call is logged. On the tool side, only approved MCP servers are reachable, access is scoped by role, credentials live in one place instead of in plaintext configs, and every tool call lands in an audit log.
What you get
- Model gateway: keys and budgets per team, an allowed-models list, and request logs
- MCP gateway: an approved catalog of MCP servers for your tools, scoped by role
- Central credentials: API tokens moved out of local config files
- Audit log of model and tool calls, sent to your existing logging or security stack
- SSO in front of the gateway
- Client configs for Claude Code, Cursor and other tools pointing at the gateway
Tools / stack
- LiteLLM: model gateway (keys, budgets, routing, logs)
- Obot or IBM ContextForge: MCP gateway
- Okta / Entra ID: single sign-on
- OpenTelemetry: into Datadog, Splunk or your stack
When it fits
AI use is spreading across the company faster than anyone can track, and security or finance want visibility and control without blocking it. The Plugin Registry distributes your tools; the AI Gateway enforces who can use them at runtime.
Add a Workshop (+$500): your engineers watch us install it live and learn the patterns alongside the product. 4 hours, real work.
Start with a 30-minute call
Free 30-minute discovery call. We'll see whether this fits your team.