OpenAI Dots vs AWS Bedrock AgentCore

OpenAI Dots is the better fit for individuals who want an always-on personal agent inside ChatGPT; AWS Bedrock AgentCore is the better fit for engineering teams that want modular, AWS-native services to build and run their own agents. They serve different audiences, and each ties you to one vendor: Dots to OpenAI, AgentCore to AWS. Runtime is the AI agent harness for payment and fintech teams, a shared, governed place for many teams to run agents across clouds and model providers.

People compare OpenAI Dots and AgentCore because both give an agent an isolated computer in the cloud and run long tasks in the background. Dots is an agent you switch on. AgentCore is infrastructure you build on.

OpenAI Dots vs AgentCore at a glance

OpenAI Dots are always-on personal agents in ChatGPT, launched at OpenAI DevDay on September 29, 2026. Each Dot runs on GPT-6 Astra with its own computer and browser in the cloud, keeps working toward your goals between conversations, uses the plugins you connect (more than 4,000 apps), and reaches you in ChatGPT, by voice, in Slack, or in Microsoft Teams when it needs a decision.

Amazon Bedrock AgentCore is AWS's agentic platform for building, deploying, and operating agents with any framework and any foundation model. It is a set of modular services you can use together or alone: Harness (a managed agent loop, generally available since June 17, 2026), Runtime, Memory, Gateway, Identity, Code Interpreter, Browser, Observability, Evaluations, Optimization, Policy, Registry, and Payments.

Audience and product scope

OpenAI Dots. A finished product for one person. You give it a goal, and it researches, analyzes data, prepares documents, or builds software, then checks in when it needs you. At launch each eligible user gets one Dot, and its memory is built from that user's conversations, ChatGPT memory, and its own notes.

AgentCore. Infrastructure for engineers. There is no end-user interface: you decide how the services fit together, write or configure the agent logic, and build the product people use. Harness lets you declare the model, prompt, tools, and memory in configuration. If you want your own loop, Runtime hosts agents built with LangGraph, CrewAI, OpenAI Agents SDK, Strands, and others.

Controls and data

OpenAI Dots. Built-in safeguards and an automatic review check actions that affect your accounts or share information. Each user can set custom rules: act without asking, act when told, ask first, or hand off. A monitoring system can pause or stop a Dot. In Enterprise, Dots are off until an admin enables them and sets who can use them.

AgentCore. Policy intercepts every tool call at the Gateway and checks it against rules written in natural language or a Cedar-compatible policy language. Identity works with Cognito, Okta, Entra ID, and Auth0. Observability uses OpenTelemetry and CloudWatch, inside your existing AWS controls. Which actions need a human, who approves them, and where the request shows up are design decisions your team still builds.

Models and infrastructure

OpenAI Dots. OpenAI's model on OpenAI's cloud computer, inside ChatGPT. The tight integration is the point; there is no model choice or bring-your-own infrastructure.

AgentCore. Model-agnostic: Harness works with Bedrock models, OpenAI, Google Gemini, and OpenAI-compatible providers, and AWS says you can switch model providers mid-session without losing context. Runtime works with any model, including Claude, Nova, Llama, and Mistral. Each session runs in its own microVM. The managed services run in AWS regions.

So the two lock-ins point in different directions. Dots fixes the model and the cloud. AgentCore leaves the model open and fixes the cloud.

Pricing

OpenAI Dots. One Dot is included with ChatGPT Pro and Business Premium at no extra cost, with an Enterprise beta. OpenAI plans to sell additional Dots and more capacity later; those prices are not published.

AgentCore. Consumption-based with no upfront commitment. Runtime microVMs are $0.0895 per vCPU-hour and $0.00945 per GB-hour on v1, or $0.1276 and $0.0169 on v2, billed per second. Gateway is $0.005 per 1,000 invocations and Policy is $0.000025 per authorization request. Model usage is billed on top. The engineering time to build approvals, audit export, and ops-facing interfaces is not on the price list.

Where Runtime enters the picture

Runtime is the AI agent harness for payment and fintech teams. Dots helps one person on OpenAI's stack; AgentCore gives engineers parts to build with on AWS. Neither is a governed place where payment ops, risk, compliance, finance, and support run shared agents on money without a build project first.

Take a sponsor bank asking why a batch of payouts settled late. The answer spans the ledger, processor logs, bank files, and the support tickets that came in, and compliance approves the reply before it goes out. A Dot can help the one person who owns it. On AgentCore you can build that workflow, after you build the approval flow, the analyst interface, and an audit export your sponsor bank accepts, which can take an engineering team two quarters. On Runtime it is one investigation with one record, worked by shared agents, with a person approving before anything is sent.

• No single vendor to depend on. Multi-cloud (your AWS, GCP, or Azure account, or self-hosted), multi-sandbox with fallbacks, and multi-model across Claude, GPT, Gemini, and open-weight models, with harness routing across Claude Code, Codex, and OpenCode. If a provider has an outage, changes pricing, or deprecates a model, agents keep running.
• Guardrails you set. Read-only by default, approvals before money moves, RBAC, masked credentials, card numbers and SSNs stripped from prompts and logs, and a full record of every run for your sponsor bank.
• Built for teams. Shared agents, templates, and one memory across support, payment ops, finance, risk, and compliance, rather than one agent per person or one stack per team.
• An engineer, not an account executive. A forward-deployed AI engineer who built payment systems at Hulu's payments team at Disney, Finix, Modern Treasury, or BlackRock builds the first agents with you. Rain saved $750k in vendor and salary spend with Runtime, which it uses across technical support, fraud and risk monitoring, compliance, cybersecurity, engineering investigations, PR reviews, QA testing, and reporting.

Turn your SOPs into agentsConnect your tools, even legacy onesKeep a human in the loopNo single vendor to depend onOne memory across teamsEvery run is stored

[::]New agent

Create a banking ops agent that handles RFIs, hold-harmless requests, and returns of funds from our partner bank, verifies against our ledger, and drafts replies for my approval.

Describe the agent you want

Agent ready

bank-ops-agent

Tools picked from your prompt

EmailLedgerSFTPMastercard

Which should you choose

• Choose OpenAI Dots if you want a personal agent for research, writing, and analysis, and your team already lives in ChatGPT.
• Choose AgentCore if you have a platform team that wants to own agent infrastructure on AWS, with model choice, a Cedar-compatible policy engine, and everything inside your AWS controls.
• Choose Runtime if payment ops, risk, compliance, or finance need shared agents on real queues, with guardrails on money movement and no dependency on one cloud, sandbox, or model vendor.

For more head-to-heads, see OpenAI Dots vs Claude Managed Agents, Claude Managed Agents vs AWS Bedrock AgentCore, and LangChain vs AWS Bedrock AgentCore.

Frequently asked questions

Is OpenAI Dots a competitor to AWS Bedrock AgentCore?

Not directly. Dots is an end-user product: a personal agent inside ChatGPT. AgentCore is a developer platform: a set of AWS services for building, deploying, and operating agents. They are compared because both give an agent an isolated computer in the cloud and let it work on long tasks without someone watching.

Can I run OpenAI models on AgentCore?

Yes. AWS says AgentCore Harness works with Amazon Bedrock models, OpenAI, Google Gemini, and OpenAI-compatible providers, and AgentCore Runtime works with any foundation model. That is OpenAI models inside AWS, not the Dots product itself.

Which is better for building an agent into our product?

AgentCore. It gives engineers a managed agent loop, per-session microVMs, memory, identity, a tool gateway, and a policy engine. Dots is a personal agent for the ChatGPT user who owns it.

Which is better for personal productivity?

Dots. It works out of the box in ChatGPT, Slack, and Teams, keeps working between conversations, and uses the apps you connect. AgentCore needs engineering work before anyone can use it.

Which is cheaper, OpenAI Dots or AgentCore?

They are priced for different buyers. One Dot is included with ChatGPT Pro and Business Premium, and OpenAI has not published prices for more. AgentCore meters each service separately, for example Runtime microVMs by vCPU-hour and GB-hour, with model usage billed on top.

Where does Runtime fit next to OpenAI Dots and AgentCore?

Dots helps one person on OpenAI's stack, and AgentCore gives engineers parts to build with on AWS. Runtime is the harness where payment ops, risk, compliance, finance, and support run shared agents on real queues, with approvals before money moves, an audit trail, and routing across Claude, GPT, Gemini, and open-weight models in AWS, GCP, Azure, or self-hosted.