Search overlay panel for performing site-wide searches

Boost Performance & Scale with Postgres Advanced. Join Pilot Now!

Heroku Managed Inference and Agents

Simplify AI in your applications through managed inference, agents, and the power of an AI PaaS

Get Started Now

AI integration made easy with Managed Inference and Agents

Heroku Managed Inference and Agents provides a managed inference service, delivering powerful models and tools as a trusted, secure, and scalable service optimized for developers. Simplify your AI operations and focus on building your application, not managing infrastructure.

View Documentation Heroku Managed Inference and Agents Plans

Why Heroku Managed Inference and Agents?

Leading AI models, managed for you

Access an opinionated set of models from the world’s top AI providers, chosen for their generative power and performance. Whether you’re exploring AI for the first time or building enterprise-grade solutions, Heroku Managed Inference and Agents provides the right models optimized for ease of use and efficacy in the domains you need most.

Performance and trust

Benefit from Heroku’s world-class infrastructure that delivers high performance and secure AI services. We handle the complexity and overhead of operating AI infrastructure and systems, so you can focus on your core business needs with confidence.

Graceful developer experience

Adding AI to your application is as simple as attaching the Heroku Managed Inference and Agents add-on to your app or using heroku ai:models:create in the CLI. Test and evaluate models directly from the command line with heroku ai:models:call, making it easy to optimize prompts and debug interactions without complex setups.

Seamless application integration

Heroku Managed Inference and Agents provides environment variables for selected models, making it seamless to call from within your application. Config vars make it easy to securely connect to your Heroku Managed Inference and Agents endpoints, so you can focus on building features rather than managing infrastructure.

Powerful agentic capabilities

Create AI applications that can take action and interact with your systems. Using MCP with Heroku Managed Inference and Agents enables agents to access external tools, manipulate data, and execute functions within your application. Design AI assistants that can analyze spreadsheets, query databases, call APIs, and generate reports – all while respecting your security boundaries and business logic.

Focus on value, not complexity

Just as Heroku simplified application deployment and scaling, Heroku Managed Inference and Agents makes AI accessible so you can focus on the value of your AI-enhanced applications rather than taking on the complexity of operating rapidly evolving AI technology.

Terminal window showing commands and output for creating a Heroku app called "mia-app" with a Claude AI model add-on, including provisioning status and setup instructions.

Get started now with Heroku Managed Inference and Agents

  1. Create a Heroku app
  2. Attach Heroku Managed Inference and Agents add-on to your app
  3. Access top AI models
  4. Deploy your code to your app
  5. Build AI-powered experiences

View Docs

Build intelligent apps

Intelligence at the core

Embed AI capabilities directly into your application’s core logic. Heroku Managed Inference and Agents enables you to integrate AI models for text generation, reasoning, and content creation with minimal overhead. Build apps that adapt to user needs, generate dynamic content, and deliver personalized experiences with high performance and predictable pricing.

Multi-model, multi-purpose

Access a curated selection of leading AI models tailored for different needs and budgets. Choose language models optimized for speed, quality, or cost. Heroku Managed Inference and Agents’ unified access pattern makes it easy to switch between models or use different models for different parts of your application, all while maintaining a consistent developer experience.

Agentic applications

Create AI applications that can take action and interact with your systems. Using MCP with Heroku Managed Inference and Agents enables agents to access external tools, manipulate data, and execute functions within your application. Design AI assistants that can analyze spreadsheets, query databases, call APIs, and generate reports – all while respecting your security boundaries and business logic.

Heroku Managed Inference and Agents models

Claude Haiku 4.5

A fast and cost-effective model optimized for high-throughput tasks and real-time interactions.

Text-to-Text
  • Fast inference
  • Cost-effective
  • Resource efficient
  • Tool calling (MCP)

Claude Opus 4.5

A next-generation frontier model offering superior reasoning and nuance for complex tasks.

Text-to-Text
  • Complex reasoning
  • Agentic workflows
  • Coding & research
  • Tool calling (MCP)

Claude Sonnet 4.5

A high-performance model that balances intelligence and speed for complex tasks.

Text-to-Text
  • High accuracy
  • Complex reasoning
  • Creative generation
  • Tool calling (MCP)

gpt-oss-120b

An open-weight model for high-reasoning and agentic tasks.

Text-to-Text
  • General purpose
  • Cost-effective
  • Open-weight model
  • Tool calling (MCP)

Kimi K2 Thinking

A specialized reasoning model designed for chain-of-thought applications and complex logic puzzles.

Text-to-Text
  • Complex reasoning
  • Math and logic
  • Writing
  • Tool calling (MCP)

MiniMax M2

A highly efficient general-purpose model with competitive pricing structure.

Text-to-Text
  • Cost-effective
  • Interleaved thinking
  • Creative generation
  • Tool calling (MCP)

Nova Pro

A model optimized for complex scenarios requiring high accuracy and deep reasoning.

Text-to-Text
  • General purpose
  • Deep reasoning
  • Complex, context-heavy tasks
  • Tool calling (MCP)

Nova 2 Lite

A model built for speed, delivering rapid, low-cost inference for real-time applications.

Text-to-Text
  • Cost-effective
  • Fast inference
  • Everyday tasks
  • Tool calling (MCP)

Qwen3 235B

A high-parameter open-weight models optimized for enterprise deployment.

Text-to-Text
  • Hybrid modes
  • Complex reasoning
  • Multilingual
  • Tool calling (MCP)

Qwen3 Coder 480B

A high-parameter open-weight models optimized for enterprise deployment.

Text-to-Text
  • Agentic coding
  • Complex reasoning
  • General purpose
  • Tool calling (MCP)

Cohere Embed Multilingual

A state-of-the-art embedding model that supports multiple languages.

Text-to-Embedding
  • Semantic search
  • Text similarity
  • High-dimensional vectors

Stable Image Ultra

A state-of-the-art diffusion (image generation) model.

Text-to-Image
  • High-quality images
  • Creative generation
  • Customizable outputs

Ready to get started with Heroku Managed Inference and Agents?

View Heroku Managed Inference and Agents Plans