Console
Cognipeer Console
An open-source, self-hosted AI control plane. One OpenAI-compatible endpoint in front of every provider you use, with the knowledge engine, guardrails, tracing, and cost controls a team needs to run models in production.
- Licence AGPL-3.0 community edition
- Deploy Self-hosted
- API OpenAI-compatible
Console is documented for two readers. If you are running the control plane — deploying it, connecting providers, setting policy, watching spend — start with the Guide. If you are building against it, the Developer section covers the architecture, the core modules, and the full API reference.
Get started
Install Console, connect your first provider, create a model, and route a client to it.
How-to guides
Task-shaped walkthroughs: guardrails, tracing, crawling, browser automation, token optimization.
Architecture
How the control plane is put together, and how the core modules fit around the request path.
API reference
Every endpoint — chat, embeddings, vectors, agents, guardrails, tracing, spend, and more.
Release notes
What shipped in the 1.0, 1.1, and 1.2 community lines, and in the enterprise releases.
Licensing
What the AGPL-3.0 community edition covers, and where commercial licensing applies.
What Console does
Routes every model call
One OpenAI-compatible endpoint in front of hosted providers and your own GPU fleet, with dynamic routing, fallbacks, and per-tenant budgets.
Grounds answers in your data
A knowledge engine with crawling, chunking, vector stores, and reranking, so retrieval is part of the control plane rather than bolted onto each app.
Keeps it governed
Guardrails and PII policies on the way in and out, agent tracing for what actually happened, and spend attribution for what it cost.
Where to go next
- Getting Started — install, first provider, first model.
- Configuration — every environment variable, with defaults.
- Architecture — the request path and the module layout.
- Using the SDK — when to reach for
@cognipeer/console-sdkinstead of raw HTTP.