Integration

Connect your AI and what it reaches

Route model calls and data connections through Burgus in your environment. Same keys, same SDKs — plus active blocking, security alerts, and audit on live production traffic.

Two channels Burgus secures

Agents call models and reach data through MCP. Burgus secures both on production traffic — route through hostnames on your deployment, keep your credentials, gain security and audit.

LLM data plane

https://{provider}.securellm.{your-domain}

Prompts and completions. Same wire format and provider auth. 40+ supported providers.

MCP data plane

https://{source}.mcpdata.{your-domain}

Tool calls and data ingress/egress. Pack sources or bring your own MCP endpoint.

Control plane

Operator MCP & REST

Keys, access policies, alerts, and exports — programmatic control on your deployment.

LLM access — any provider your agents need

Register upstream keys for Groq, OpenAI, Anthropic, Gemini, Mistral, and dozens more via the operator API or MCP. Point your provider SDK base_url at the secure hostname assigned on your deployment. Your commercial terms stay with each vendor; Burgus adds active blocking, security alerts, audit, and multi-agent correlation on every call.

Optional X-Llmproxy-* headers (session, workflow, agent id) unlock enhanced multi-agent graphs — see multi-agent security.

MCP access — any data source your agents use

Activate MCP sources from the platform catalog or connect your own MCP server. Agents reach sources through *.mcpdata.{your-domain} with the client credentials those services already issued to you. Burgus audits tool arguments and results on the data plane; operator API keys are for the control plane only, never sent upstream to third-party servers.

Identity and access

Configure Access on each connection: which credentials, agent ids, and workflow ids may use a given LLM path or MCP source. As your agent fleet grows, Burgus gives you a single place to define trust boundaries and prove who touched what in production.

One audit trail across your stack

LLM and MCP traffic feed the same audit store, alert pipelines, and export surfaces. Pull combined security reports via operator MCP or REST, and export assessor-ready bundles — compliance & exports.

Getting started

License and deploy Burgus in your environment
Register upstream LLM keys — copy base_url hostnames from your deployment
Activate MCP sources and register client credentials; set Access policies
Route production agent traffic through assigned hostnames
Configure blocking policies on your deployment
Review blocks, alerts, and exports via operator MCP or REST

Can we use different providers for different agents?

Yes. Burgus is built for multi-vendor production. Register the keys you use and route each agent through the secure hostname that matches your stack.

Do we keep our existing provider contracts?

Yes. BYOK — model and MCP fees stay on your direct invoices with each vendor.

Is integration a replatform?

No. Keep your SDKs and credentials. Change hostnames to your Burgus deployment URLs.

Does Burgus secure MCP tool traffic?

Yes. The mcpdata plane mirrors the LLM integration story: hostname change, client credentials, Access configuration, full audit and alerts on tool ingress and egress.