The enterprise AI platform
built for engineering teams.
One workspace URL for your engineering team. MCP servers with zero-trust secrets, per-user analytics, and instant access revocation. No API keys on developer machines. Full audit trail on every tool call.
Four pillars. Zero boilerplate.
AI Gateway
Multi-model routing to OpenAI, Anthropic, Gemini, Azure OpenAI, and Cloudflare AI from a single endpoint. BYOK, fallback chains, per-consumer token wallets, and metered billing built in.
MCP Workspace
Aggregate N MCP servers behind one authenticated URL. Auto-namespace tools, inject team secrets, and expose your private registry to Claude, ChatGPT, Cursor, or any MCP-compatible client.
Real-Time Infrastructure
WebSocket pub/sub, presence tracking, and database change streaming — all on Cloudflare's edge. Enterprise plans include 25K concurrent connections with automatic hibernation and scaling.
Developer Auth
Email OTP, magic link, BYO-JWT (connect your existing IdP), refresh token rotation, Turnstile CAPTCHA, and per-project end-user lifecycle management — all configurable without writing auth code.
Route any model.
Bill any way.
One endpoint. Any LLM. Your keys. Enterprise teams deploy AI APIs with multi-provider fallback chains, per-consumer rate limits, and metered billing — all without touching infrastructure.
Works with
import Aerostack from '@aerostack/node';
const client = new Aerostack({
projectSlug: 'my-project',
apiKey: process.env.AEROSTACK_API_KEY,
});
const stream = await client.gateway.chat({
gatewaySlug: 'my-ai-api',
consumerKey: 'ask_live_xxxxxxxxxxxx',
messages: [{ role: 'user', content: 'Hello!' }],
stream: true,
});
for await (const chunk of stream) {
process.stdout.write(chunk.content ?? '');
} One gateway. Every tool.
AI Clients
MCP Workspace
gateway.aerostack.dev/ws/{workspace-slug}
Single authenticated URL · Tool namespacing · BYOK secrets
Your Internal Tools
Your team's private tool registry — aggregate any number of MCP servers behind a single authenticated endpoint. Every AI model your team uses can reach the same unified toolset.
- Private by default — invisible to the public marketplace
- BYOK secret injection — your internal API keys never leave your config
- Automatic tool namespacing — prevent conflicts across servers
- Fan-out tools/list in parallel — fast cold start across N servers
- Works with Claude, ChatGPT, Cursor, and any MCP-compatible client
MCP Governance — control every tool call by policy.
MCP governance is the control layer that decides which AI tools a team, agent, or token can call. Aerostack's policy engine enforces it directly at the workspace gateway — allow and deny rules, IP restrictions, and rate caps applied to every tool call before it reaches your servers.
- Tool allow / deny lists with wildcard globs (e.g. github__*)
- IP allowlists enforced at the CIDR level
- Per-day rate caps at the workspace or token level
- Per-token tool scoping — a token can only call the tools it was issued for
- Priority-ordered enforcement — the most specific rule wins
Five built-in policy templates
Glob-deny destructive verbs — delete, drop, destroy, truncate, wipe, purge, revoke.
Block critical operations and restrict high-risk deploy, publish, and execute tools.
Glob-allow read-only verbs only — get, list, read, search, fetch, describe.
Deny outbound messaging tools — email, Slack, SMS, and webhooks.
Block destructive ops with a hard rate cap of 1,000 calls per day.
The right people. The right tools.
Admin
Manage MCP servers, secrets, and tokens. Invite and remove members, assign roles, and set policies.
Member
Call tools and create their own tokens. Cannot manage servers, secrets, or other members.
Viewer
List available tools only. No tool calls, no token creation, no configuration access.
Admins invite teammates by email with a pre-assigned role and track pending invites with resend and cancel. Per-member analytics — tool call count, token count, and last active time — are visible to workspace admins. Removing a member instantly revokes every token they hold.
Security is not a feature. It's the foundation.
Single sign-on. Your IdP, your rules.
Connect your existing identity provider and let your team sign in through the directory you already manage. Enterprise supports SAML 2.0 single sign-on with Okta, Azure AD, and other providers, plus BYO-JWT to wire in your own IdP directly. Sign-in runs through a secure code-exchange flow with mandatory refresh token rotation.
One identity, every workspace
Provision and deprovision through your IdP — access follows your directory, not a separate password list.
Know exactly what's happening.
17 Usage Metrics
API requests, database queries, cache ops, storage, AI tokens, realtime connections — all tracked automatically, per project.
Anomaly Detection
Automatic abuse scoring (0–100). Score above 80 triggers burst throttle. Score above 100 flags for review.
Full Audit Trail
Tiered retention — up to 1 year on Business and Enterprise — on every user action. Entity creation, updates, and deletions, each with actor, timestamp, and diff.
5-Tier Logging
Edge logs → platform errors → developer traces → analytics events → usage rollups. Five levels, all automatic.
Per-Request Latency
P50 / P95 / P99 tracked per route and gateway API. Error rates broken down by status code.
Webhook Execution Logs
Full request/response bodies logged for every hook call. Status, latency, retry history all visible in dashboard.
Same DX.
Any scale.
Enterprise teams get the same CLI, SDKs, and type-safe RPC client as the rest of the platform. No new tooling to learn, no lock-in. Just higher limits.
SDK Languages
CLI
npx aerostack create npx aerostack deploy npx aerostack add <slug> import aerostack
client = aerostack.Client(
project_slug="my-project",
api_key=os.environ["AEROSTACK_API_KEY"],
)
# Streaming response — works across all model providers
with client.gateway.chat(
gateway_slug="my-ai-api",
consumer_key="ask_live_xxxxxxxxxxxx",
messages=[{"role": "user", "content": "Hello!"}],
stream=True,
) as stream:
for chunk in stream:
print(chunk.content or "", end="", flush=True) Enterprise vs Business
How Enterprise works.
Talk to Us
Tell us your team size, AI stack, and compliance requirements.
Custom Setup
We configure your private MCP workspace, limits, and dedicated support.
Ship in Days
Same CLI, same SDKs — enterprise-grade from day one.
Ready to deploy your
AI infrastructure?
Talk to our team about custom limits, private MCP workspaces, and dedicated support.
Custom pricing starting from Business tier ($149/mo). Volume discounts and annual invoicing available.
hello@aerostack.devRelated features