Overview

Deeplinq is an AI gateway and application engine for multi-tenant products. It exposes familiar OpenAI- and Anthropic-compatible endpoints while owning the governance layer around them: identity, model access, metering, retrieval, durable agents, personal connections, guardrails, and audit evidence.

Start here

Install Deeplinq

Run the local stack or prepare a production deployment.

Initialize the platform

Configure providers, pricing, and your first organization.

Operate the platform

Manage organizations, connectors, models, credits, and policy.

Build an application

Make requests, add knowledge, persist conversations, and run agents.

Browse the API

Find the shipped endpoint inventory and generated schemas.

What you can build

Compatible model access through /v1/chat/completions, /v1/messages, and /v1/models.
Grounded applications using tenant-scoped datasets, projects, and screened retrieval.
Durable agents with bounded multi-turn execution, tool calls, human approval, and run budgets.
Personal-data workflows through governed Gmail, Outlook, and external MCP connections.
Application workflows through projects, persistent conversations, semantic memory, teams, schedules, and signed webhooks.
Regulated product controls including model grants, spend limits, guardrails, append-only billing, and tamper-evident audit trails.

Core principle

Your application remains the identity authority for its users. Deeplinq validates the identity and policy context supplied by your trusted backend, then applies tenant isolation and governance before any provider call.

Deeplinq credentials are server-side credentials. Never expose an engine API key or tenant-signing key in browser or mobile application code.