Connect Any Provider, Framework, or Observability Pipeline
AgentMink provides unified proxy routing across all frontier providers, open-weights inference engines, developer frameworks, and webhook destinations with zero code refactoring.
OpenAI
Full compatibility with GPT-4o, GPT-4o mini, o1, and o3-mini models via direct BYOK key execution with zero token markup.
Anthropic
Direct proxy access to Claude 3.5 Sonnet, Claude 3.5 Haiku, and Claude 3 Opus with sub-millisecond format translation.
Google Gemini
High-throughput routing to Gemini 2.0 Flash, Gemini 1.5 Pro, and Flash-Lite models with cost-first arbitrage.
Groq LPU
Hardware-accelerated inference for Llama 3.3 70B, DeepSeek R1, and Mixtral delivering sub-50ms Time-To-First-Token.
DeepSeek
DeepSeek-V3 and DeepSeek-R1 reasoning models routed dynamically for complex code generation and mathematical analysis.
Mistral AI
Mistral Large 2, Codestral, and Pixtral endpoints accessible directly with automatic JSON format validation.
Together AI
Access 100+ open-weights models including Llama-3, Qwen 2.5, and specialized fine-tunes through one unified endpoint.
OpenRouter
Secondary failover backend connecting to 200+ global model endpoints with real-time rate tracking.
AWS Bedrock
Enterprise VPC and IAM credential proxying for Claude, Amazon Titan, and Jurassic models deployed in private clouds.
Vercel AI SDK
Drop-in compatibility with Next.js, Nuxt, and SvelteKit apps using createOpenAI({ baseURL }).
LangChain & LangGraph
Compatible with Python and TypeScript LangChain ChatOpenAI classes for multi-agent chains and state graphs.
openai_api_base
LlamaIndex
High-throughput RAG and document vector ingestion pipelines with cost-first extraction and quality-first synthesis.
CrewAI & AutoGen
Autonomous multi-agent orchestration frameworks protected by per-key spending caps and rate limit failover.
Slack Incident Webhooks
Instant alerts dispatched to your engineering Slack channels on spend cap thresholds, 5xx provider failovers, or quota spikes.
Discord Webhooks
Real-time community and developer notification channels for prompt volume anomalies and daily cost summary digests.
Custom HTTP Webhooks
Stream per-request telemetry payloads to your internal Datadog, Grafana, or logging microservices via secure POST endpoints.
Don’t See Your Custom or Self-Hosted Model?
AgentMink routes to any OpenAI-compatible API endpoint or self-hosted vLLM / Ollama instance. Configure private upstreams in your workspace settings.
Configure Custom Upstream →Connect Your First Provider in Under 2 Minutes
Sign up for a free developer account and start executing requests with automatic cost arbitrage, smart failover, and sub-50ms latency.