enterpilot, Inc.
enterpilot, Inc. develops GoModel, an open-source, MIT-licensed AI gateway written in Go that provides a unified OpenAI- and Anthropic-compatible API for routing requests across 37 providers. The gateway handles load balancing, failover, caching, budgets, audit logs, and usage tracking, and is designed to run as a single lightweight binary with an embedded admin UI.
- Artificial Intelligence
- Developer Tools
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Applications that rely on AI models often become tightly coupled to a single provider, making vendor switches a code-level project. Additionally, teams struggle to track costs across different users and features, debug incidents without proper logs, and avoid paying repeatedly for identical requests, all while managing provider-specific APIs and behaviors.
Solution
GoModel is an open-source, MIT-licensed AI gateway that centralizes provider routing, governance, and observability behind a single OpenAI- and Anthropic-compatible API. It authenticates each request, applies scoped workflows such as caching, budgets, and guardrails, and routes traffic to any of 37 supported providers with automatic failover and load balancing. The gateway is distributed as a single Go binary with an embedded admin UI for managing keys, budgets, and logs, and it stores data in SQLite, PostgreSQL, or MongoDB. By changing only the base URL, applications can keep their existing SDKs while gaining provider flexibility, cost control, and full request traceability.
Target Audience
Primary users are engineering teams and platform organizations building AI-powered applications who need to manage multiple model providers, control costs, and maintain observability without rewriting their application code.
Features
- Unified OpenAI- and Anthropic-compatible endpoints supporting chat, embeddings, Responses API, audio, and realtime speech
- Routing across 37 providers with weighted round-robin load balancing, cost-based routing, automatic failover, and circuit breaker protection
- Scoped workflows enabling per-path configuration of caching, audit logging, budget limits, guardrails, and rate limits
- Exact-match and semantic response caching (Qdrant, pgvector, Pinecone, or Weaviate) to reduce duplicate costs
- Virtual API keys bound to user paths and labels for team-level access control with revocation from the admin UI
- Per-request usage and cost tracking with custom pricing overrides, mapped to teams, tenants, and features
- Audit logs capturing resolved routes, provider attempts, and cache results with optional body/header logging
- Prometheus metrics and OpenTelemetry traces on GenAI semantic conventions for integration with standard observability tools
- MCP gateway aggregating multiple MCP servers behind one endpoint with namespaced tools
- Single binary deployment with Docker, Compose, and Helm recipes, plus an embedded admin dashboard
- Aliases and virtual models that publish stable names and remap the real provider and model behind them via config change
- Provider passthrough that calls any provider's native API through /p/:provider/* while retaining GoModel's auth, usage tracking, and audit