Heroku AI

What it does

Heroku AI is the AI layer of the Heroku platform. Its core is the Managed Inference and Agents add-on, which gives a Heroku app access to hosted chat, embedding and image-generation models through an OpenAI-style API. Developers send prompts or text and receive completions, vectors or images. The add-on can also let a model run tools on Heroku, including code execution and custom tools, and returns action traces in the output. pgvector on Heroku Postgres stores embeddings for similarity search.

In practice, you create and attach a model resource to an app with CLI commands, and the add-on sets config variables holding an endpoint URL and key. Your code calls those endpoints with existing OpenAI-compatible SDKs, or you test from the Heroku CLI AI plugin. Models are hosted on Amazon Bedrock, and MCP toolkits let agents such as Claude Desktop or Cursor reach servers deployed on Heroku. Its distinguishing point is that inference, agents, vector storage and deployment sit inside one Heroku workflow with Pipelines and Review Apps.

It suits product teams, engineers and startups that already deploy on Heroku or want managed infrastructure. Use requires a Heroku account and an app, and the add-on bills by token usage. Requests are rate limited per minute, long requests should stream to avoid timeouts, and model output can be inaccurate or vulnerable to prompt injection, so Heroku advises treating it as untrusted input. Model availability is limited to a curated set per region.

What it can do

• Call hosted chat, embedding and image-generation models through an OpenAI-style API
• Let models run built-in or custom tools on Heroku dynos with retries and action traces
• Deploy and manage MCP servers through Heroku MCP Toolkits
• Store embeddings and run semantic queries with pgvector on Heroku Postgres
• Monitor token usage and rate-limit headers per model resource
• Automatic prompt caching for system prompts and tool definitions, which can be disabled per request

Works with

• Amazon Bedrock
• Claude Desktop
• Cursor
• Heroku Postgres (pgvector)
• OpenAI-compatible SDKs

Example workflows

Add a chat model to an existing app

Input: a Heroku app and a chosen chat model. Steps: install the Heroku AI CLI plugin, run addons:create to attach a model resource, read the INFERENCE_URL and INFERENCE_KEY config variables, and call the chat completions endpoint from app code. Output: model responses inside the app.

Build retrieval over documents

Input: text documents. Steps: attach an embedding model, generate embeddings through the embeddings endpoint, store them in Heroku Postgres with pgvector, and query by similarity. Output: relevant passages to supply to a chat model.

Before you choose

• Billing is per token or per query and per image, so costs vary by model; check the rate table before choosing
• Streaming is recommended because requests over 29 seconds can time out, and tool calls of 55 seconds or more time out
• Apps can by default only provision models in their own region (us or eu)

Listed under

• Coding
• MCP and agent tools
• AI
• AI agents
• AI paas
• App deployment
• Developers
• Managed inference

Get started

Create an app, then provision the add-on with the Heroku CLI

1. Sign up for Heroku and create an app
2. Install the Heroku CLI and the @heroku/plugin-ai plugin
3. Run heroku addons:create heroku-inference:standard for your app and call the model with the generated config variables

The add-on bills by usage; detached or unused model resources do not bill. See the Heroku AI pricing table for per-model rates.

Source: heroku.com/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Build and Scale AI Apps Easily with Heroku AI ↗www.heroku.com
• Managed Inference and Agents Add-on | Heroku Dev Center ↗devcenter.heroku.com
• Cloud Application Platform Pricing & Plans | Heroku ↗www.heroku.com
• Cloud Platform for Developers | Build & Ship Faster | Heroku ↗www.heroku.com

Company website

heroku.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.